See

discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/44453940

Linguistic communication and speech acts / Kent Bach, Robert M.

Harnish

Article
Source: OAI

CITATIONS READS

3 1,001

2 authors, including:

Kent Bach
San Francisco State University
71 PUBLICATIONS 3,394 CITATIONS

SEE PROFILE

All content following this page was uploaded by Kent Bach on 13 August 2017.

The user has requested enhancement of the downloaded file.

~inguistic Kent..Bach
Communication and RobertM. Harnish
Speech Acts

The MIT Press

Cambridge, Massachusetts, and
London, England
 Contents

Preface xi 2.2. Inferring Operative

Meaning 20
Introduction xiii
2.3. Inferring Locutionary
Acts 24
Part One 204. Saying That 29
The Theory 2.5. Determining Literal
Illocutionary Force
Chapter One Potential from Locutionary
Linguistic Communication: Acts 34
A Schema for Speech Acts 3
2.6. Conclusion 37
1.1. Components of Speech
Acts 3 Chapter Three
1.2. Simple Version of the A Taxonomy of Communicative
Speech Act Schema (SAS) 4 Illocutionary Acts 39
1.3. Saying and the 3.1. Constatives 42
Linguistic Presumption 8 3.2. Directives 47
104. Literal Illocutionary Acts 10 . 3.3. Commissives 49
1.5. The Communicative 3 A. Acknowledgments 51
Presumption and
Illocutionary Intentions 12 . 3.5. Felicity Conditions 55
1.6. Illocutionary Intentions 3.6. Infelicity and Obvious
and Effects 15 Insincerity 57
1.7. Perlocutionary Acts and Chapter Four
Effects 16 Elaborating the Schema:
Illocutionary Acts 60
Chapter Two
Elaborating the Schema: 4.1. Literal (and
Locutionary Acts 19 Direct) Illocutionary
Acts 60
2.1. Austin on Locutionary
Acts 19
Contents viii Contents ix

4.2. Nonliteral Illocutionary 7.2. Why Communicative Chapter Ten

Acts 65 Illocutionary Acts Are Not Standardization and
Conventional 122 Dlocutionary Devices 203
4.3. Indirect Illocutionary
Acts 70 7.3. Constitutive Rules and 10.1. Performatives as
Communication 127 Constatives 203
4.4. The Elaborated Schema 76
4.5. Perlocutionary Acts 81
7.4. Performatives and 10.2. Embedded and
Convention 131 Hedged Performatives 209
Chapter Five 7.5. Why Locutionary 10.3. Illocutionary
The Status and Scope of the Acts Are Not Fully Adverbials 219
Speech Act Schema 84 Conventional 132
10.4. Syntactic Liberties 228
5.1. The SAS and the
Analysis of Linguistic Chapter Eleven
Communication 85 Part Two
The Speech Act Schema and
Issues 234
5.2. The Explanatory Value Psychology
of the SAS 89 Chapter Eight 11.1. Production 236
5.3. On the Psychological Locutionary Acts: 11.2. Comprehension 240
Reality of the SAS 91 Philosophical and Linguistic
Issues 137
5.4. The SAS and Social
Psychology 93 8.1. Linguistic Meaning - 138 Appendix
149
Mutual Belief and Social
5.5. Collateral Acts and 8.2. Speaker Meaning 267
Concepts
Social Moves 96 8.3. Presupposition 155
I Social Regularities 270
8.4. Implicature 165 277
Chapter. Six Social Collectivities
Conventional Illocutionary 8.5. Conclusion 172
Conclusion 283
Acts 108
108 Chapter Nine 285
6.1. Conventions Notes
Indirect Acts and Illocu-
6.2. EfIectives and tionary Standardization 173
Bibliography 311
Verdictives 110
9.1. The Ambiguity
6.3. Conventional Thesis 174 Index 321
Illocutionary Acts and
9.2. The Conventionality
Institutional Facts 113
183
Thesis
6.4. Performing Conven-
tional Illocutionary Acts 116 9.3. The Standardization
Thesis 192
Chapter Seven 9.4. Infelicity and
Communication and Standardized
Convention 120 Indirection 195
7.1. Rules and 9.5. Standardization and
Conventions 120 Paragrammatic Facts 198
 Preface

Extensive work in the study of language has been stimulated by the

work of Chomsky on grammar, Grice and Katz on meaning, and Austin
and Searle on speech acts. Nevertheless, little has been done to inte-
grate these topics into a general account of linguistic communication,
even though it is widely recognized that to communicate linguistically
is more than just saying something-what is communicated is deter-
mined not merely by what is said. The structure and meaning of the
expressions used are essential, but so are the speaker's intention and
the hearer's recognition of it. In our view a communicative intention
has the peculiar feature that its fulfillment consists in its recognition.
The speaker intends the hearer to recognize the point of his utterance
not just through (1) content and (2) context but also because (3) the
point is intended to be recognized. '
The two major alternative theories, Sadock's (1974) and Searle's
(1969), fail to do justice to all three factors. Moreover, neither explains
just how linguistic structure and speech acts are connected-and how
they are not. Sadock focuses on linguistic structure and Searle on
speech acts, but both assume (in different ways) that this connection is
mostly semantic. Our view is that the connection is inferential, not just
for nonliteral and for indirect utterances but even for literal ones. The
inference is simplest in the literal case, but each case involves all three
factors: content, context, and communicative intention.
These factors require systematic explanation. Accordingly, we at-
tempt to characterize precisely the nature of communicative intentions
together with the nature of the inference the hearer makes in identify-
ing them. Within the framework of this account we offer a detailed
classification of speech acts and discuss the function of linguistic de-
vices and the role of social conventions in the performance of speech
Preface xii
acts. Although our approach is primarily philosophical and linguistic,
it intersects with cognitive and social psychology by exploring psy-
chological and social factors that contribute to successful linguistic
communication.
Our aim is ambitious: to present a conception of linguistic communi-
cation that integrates philosophical, linguistic, and psychological is-
sues. We have profited from the intellectual legacy of the authors cited,
as well as from the work and encouragement of our friends Stephen
Schiffer, Bruce Fraser, George Spanos, and Chuck Carr. Special
thanks go to George E. Smith (for his voluminous comments), Tom
Larson (for the index), Claire and Polly Baker, and Judy Miles. Thanks
are due also to our many students who have endured cruder versions of
our theory and have helped us, either by their questions and sugges-
tions or by their noncomprehension, to produce clearer formulations of
our ideas.
Introduction
There is nothing people do more often, in more ways, than talk to one
another. For most people, nothing is easier. Sometimes we have to
struggle to find the right words or to get them out, sometimes we must
pause to discern or decipher what someone else says, but on the whole
we speak fluently and understand others effortlessly. Yet using lan-
guage is a very complex enterprise, as anyone knows who has tried to
master a foreign language. Moreover, much goes into using ~ language
besides knowing it and being able to produce and recognize sentences
in it. Exchanging words is a social affair, usually taking place within the
context of a fairly well-defined social situation. In such a context we
rely on one another to share our conception of what the situation is.
With people we know, rather than spell everything out we rely on
shared understandings to facilitate the process of communicating. What
sort of process is it? Linguistic communication is easily accomplished
but not so easily explained.
There is a popular and venerable conception of linguistic communi-
cation as conveying a message. It dates back at least to John Locke
(1691, III, i) and has been stated most recently as follows:
A has in his mind some sort of message (or idea), and he wishes B to
form in his head the same message. This message is transformed ulti-
mately into a series of neural impulses that are sent to the muscles
responsible for the actual production of speech, which follows im-
mediately ... The listener, B, must decode A' s message by converting
the sounds into a semantic representation. (Cairns and Cairns 1976,
17-18)
Even on this popular conception linguistic communication involves
more than transmitting a signal; inference on the part of the hearer is
required.
Introduction xiv
Commonsensically: Communication is successful only when the hearer
infers the speaker's intentions from the character of the utterance he
produced. (Fodor 1975, 103)
In more detailed and sophisticated form the commonsensical view goes
like this:
The speaker's message is encoded in the form of a phonetic representa-
tion of an utterance by means of the system of linguistic rules with
which the speaker is equipped. This encoding then becomes a signal to
the speaker's articulatory organs, and he vocalizes an utterance of the
proper phonetic shape. This is, in turn, picked up by the hearer's audi-
tory organs. The speech sounds that stimulate these organs are then
converted into a neural signal from which a phonetic representation
equivalent to the one into which the speaker encoded his message is
obtained. This representation is decoded into a representation of the
same message that the speaker originally chose to convey by the
hearer's equivalent system of linguistic rules. (Katz 1966, 103-104)
This sophisticated rendition of the commonsensical view is highly
plausible-as far as it goes. It captures the mechanical, context-
independent aspects of linguistic communication, but it leaves much to
be accounted for. First, utterances can be ambiguous, so disambigua-
tion must be considered. Second, a person need not speak literally and
directly. Instead of meaning just what he says, he may mean something
else or something more; to be understood, the utterance requires more
than being decoded in the way Katz describes. Third, the common-
sensical view neglects the role of shared understandings in successful
communication. Finally, it does not tell us what exactly these messages
are. Indeed, it gives the impression that linguistic communication is
limited to conveying information and ignores acts of making requests,
offering apologies, conveying greetings, and so on.
Linguistic communication is not exclusively a matter of conveying
information, that is, of making statements. J. L. Austin's pioneering
investigation (1962) of the variety of speech acts opened philosophers'
eyes to the breadth of ways in which language can be used. In present-
ing a preliminary version of our account of linguistic communication, in
chapter 1 we adopt a version of Austin's well-known distinction be-
tween locutionary, illocutionary, and perlocutionary acts by charac-
terizing the sorts of intention with which each act is performed. Of
special importance is our account of (communicative) illocutionary in-
tentions. They are reflexive intentions, in the sense of H. P. Grice
(1957): a reflexive intention is an intention that is intended to be recog-
Introduction xv
nized as intended to be recognized. We further restrict illocutionary
intentions to those intentions whose fulfillment consists in nothing
more than their recognition. The sort of reflexive intention that has this
feature is that of expressing an attitude (such as a belief or desire).
Accordingly, an act of linguistic communication is successful if the
attitude the speaker expresses is identified by the hearer by means of
recognizing the reflexive intention to express it. Of course, recognizing
that there is some such intention is not to identify the specific attitude
expressed-it is identified on the basis of what is said, together with
what we call mutual contextual beliefs.
What is said is the content of locutionary acts, the topic of chapter 2.
What the speaker says is largely, but not entirely, a matter of what his
words mean. After all, they may be ambiguous and their reference (in
the case of referring expressions) underdetermined, and so the hearer
must rely on more than just his knowledge of the language (and the
supposition that the speaker shares this knowledge) to determine what
the speaker is saying. All together, what is said depends on what ex-
pression is uttered, what meaning it has in the language, what the
speaker means by it, and what things he is referring to. We offer a
schematic account of how the hearer identifies these various items.
Once he identifies them, he has identified what is said; from that, to-
gether with mutual contextual beliefs, he can proceed to the identifica-
tion of the speaker's illocutionary act, that is, of what attitude the
speaker is expressing.
Before representing this phase of the hearer's inference, in chapter 3
we attempt to sharpen and systematize Austin's detailed classification
of illocutionary acts in order to make explicit the full range of commu-
nicative acts to which our account is meant to apply. They are cate-
gorized in terms of the kind of attitude each expresses. Then in chapter
4 we spell out the illocutionary phase of the hearer's inference. Our
schematization of this inference, which we call the speech act schema
(SAS), applies not only to literal but also to nonliteral and indirect
utterances. In chapter 5 we explain how our theory operates both as an
analysis of linguistic communication and as a representation of what it
takes (linguistically, psychologically, and socially) to communicate. In
our view, then, an act of linguistic communication is an act of express-
ing an attitude by means of saying something. What type of attitude is
expressed determines the kind of illocutionary act being performed.
That act is successful-communication has been achieved-if the
hearer identifies the attitude expressed in the way the speaker intends
Introduction xvi
him to identify it. This requires the hearer to make an inference, an
inference based partly on the supposition that the speaker intends him
to make it.
Our intention-and-inference approach contrasts sharply with Aus-
tin's view ofillocutionary acts as conventional. He neglected to explain
what he meant by conventional, much less why he considered such acts
to be conventional, but John Searle (1969) has since developed this sort
of theory, using his controversial notion of constitutive rules. Our the-
ory is thoroughly at odds with Searle's, but we do allow, following
P. F. Straws on (1964), that certain kinds of illocutionary acts involve
convention rather than intention-and-inference. Those acts are not es-
sentially communicative, however, as the examples of christening,
nominating, and acquitting illustrate. It is no coincidence that cases like
these influenced Austin's thinking about illocutionary acts in general.
In chapter 6 we characterize and categorize conventional illocutionary
acts. Chapter 7 deals with assorted topics related to convention, in-
cluding the difference between rules and conventions, Searle's theory
based on constitutive rules, conventions and explicit performatives,
and the conventionality of locutionary acts.
The seven chapters of part I present the essentials of our theory of
linguistic communication and speech acts. In part II we take up various
philosophical, linguistic, and psychological issues raised by our theory.
Chapter 8 investigates the concepts of linguistic meaning and speaker
meaning, together with the allied notions of presupposition and impli-
cation. Clarifying these notions enables us to sharpen our conception of
locutionary acts. In general, the linguistic meaning of an utterance does
not severely delimit, much less determine, the speaker's communica-
tive illocutionary intent. It does delimit what the speaker says, which in
tum provides part of the basis on which the hearer infers the speaker's
intent, but because the speaker could be speaking nonliterally or indi-
rectly, the hearer must rely on much more than the linguistic meaning
of the utterance to determine the speaker's intent. However, as investi-
gated in chapters 9 and 10, there are certain kinds of linguistic devices
that seem directly tied to the performance of particular kinds of illocu-
tionary acts. Explicit performatives, first investigated by Austin, are
the most familiar, but there are also the hedged performatives dis-
cussed by Fraser (1975) and a variety of constructions investigated by
Sadock (1974). Apart from their intrinsic linguistic interest, the rele-
vance of these devices to our theory is that their use short-circuits the
pattern of inference followed by the hearer in identifying the speaker's
Introduction xvii
illocutionary intent. This is the phenomenon, of illocutionary stan-
dardization, as we call it. Other theorists have attempted to account for
this phenomenon in terms of linguistic meaning (Sadock 1974) or illo-
cutionary convention (Searle 1975a), but we hold that short-circuited
inference is still inference-inference compressed by precedent rather
than by meaning or convention.
Our account of linguistic communication and speech acts is primarily
analytical, not empirical. To be sure, we rely on everyday experience
for the extensive variety of communicative phenomena we consider,
but the view we present is essentially a conception of what it takes to
communicate and to understand. Our view is that linguistic communi-
cation essentially involves the speaker's having a special sort of inten-
tion (an intention that the hearer make a certain sort of inference) and
the hearer's actually making that inference. In proposing this analytical
conception of linguistic communication, we could ignore empirical
questions and take the stubborn position that if in fact people do not
have the intentions and make the inferences we attribute to them, they
do not engage in communication when they talk to one another. Then
the question would arise, What do they do? We cannot demonstrate
that people do what we claim they must do if they communicate, but in
chapter 11 we try at least to make this possibility empirically plausible
in the light of current investigations in the psychology of language.
Moreover, we suggest further directions for empirical research, whose
eventual fruitfulness would be one mark in favor of our theory.
Part One The Theory
Chapter One Linguistic Communication:
A Schema for Speech Acts

People don't speak merely to exercise their vocal cords. Generally, the
reason people say what they say when they say it is to communicate
something to those they are addressing. That is, in saying something a
person has a certain intention, and the act of communicating succeeds
only if that intention is recognized by the hearer. The intention is
recognized partly on the basis of what is said, but only partly. What is
said does not fully determine what the speaker is to be taken to be
doing. If he says "I'm going to pay you back for that," he could be
making a promise or issuing a threat. How does the hearer decide
which? And how does the speaker know which way the hearer will take
his utterance?

1.1. COMPONENTS OF SPEECH ACTS

Before taking up those questions, we need to distinguish the different

aspects of a speech act. If S is the speaker, H the hearer, e an expres-
sion (typically a sentence) in language L, and C the context of utter-
ance, the main constituents of S's speech act can be schematically
represented as follows:
Utterance Act: S utters e from L to H in C.
Locutionary Act: S says to H in C that so-and-so.
Illocutionary Act: S does such-and-such in C.
Perlocutionary Act: S affects H in a certain way.1
These acts are intimately related. In uttering e, S says something to H;
in saying something to H, S does something; and by doing something, S
affects H. Moreover, the success of the perlocutionary act depends on
H's identifying one of the other acts. Our problem is to specify as
The Theory 4
Schema for Speech Acts 5

precisely as possible the nature of these acts as well as their relations to

the speaker intends to be thereby doing. However, what he says under-
one another. 2
determines what he can reasonably expect to be taken to be intending.
Clearly there is more to a speech act than saying something (per-
Suppose S says "I love you like my brother." There are various ways
forming a locutionary act), but our preliminary characterization gives
H could take this, depending on what he can infer S's intention to be
no indication of the difference between illocutionary and perlocution-
under the circumstances, given what H believes about S and in par-
ary acts. Austin's distinction in terms of what is done in saying some-
ticular what H believes S to believe H believes about S. Normally, H
thing and what is done by saying something (1962, lectures IX and X) is
can assume that if S says "I love you like my brother," S means that he
suggestive at best, since it does not explain the distinction it marks. 3
loves H as he (S) loves his own brother. But if a woman says to a man
Illocutionary and perlocutionary acts can both produce effects on the
"I love you like my brother," the man can infer (taking himself to be
he.arer, but according to Austin (p. 116) a successful illocutionary act
intended to infer) that the woman has a feeling that is more familial than
bnngs about "understanding of the meaning and of the force of the
amorous. Perhaps, however, it is not the kind of love but the amount of
locution," that is, it secures uptake. Strawson (1964a, 459) suggests
love that is in question, as, for example, where two wartime buddies
that for illocutionary acts, the effectiveness of the speaker's intention
are involved. Or "I love you like my brother" might be uttered by one
requires that the intention be recognized by the hearer: "The illocu-
man to another where it is recognized that the speaker hates his
tionary force of an utterance is essentially something that is intended to
brother. In this case H would no doubt take S as informing H that he
be understood." That is, part of the speaker's intention is that the
hates him.
hearer identify the very act the speaker intends to be performing, and
In general, the inference the hearer makes and takes himself to be
successful communication requires fulfillment of that intention.
intended to make is based not just on what the speaker says but also on
I~ general, we cannot rely on our vocabulary of verbs of speech
mutual contextual beliefs (MCBs), as we call such salient contextual
actlon to mark the distinction between illocutionary and perlocutionary
information. With the example "I love you like my brother," in one
acts. Although Austin's conception of the distinction is different from
case the crucial MCB is that the woman does not have amorous feelings
the one we wish to develop, he himself recognized that "the same word
toward her brother, whereas in another it is that the speaker hates
may genuinely be used in both illocutionary and perlocutionary ways
his brother. We call such items of information "beliefs" rather than
and that many illocutionary acts are cases of trying to do some per-
"knowledge" because they need not be true in order to figure in the
locutionary act" (1962, 145-146). For acts like ordering, warning, in-
speaker's intention and the hearer's inference. We call them "contex-
forming, and assuring, we must distinguish the ultimate perlocutionary
tual" because they are both relevant to and activated by the context of
effect the speaker is trying to achieve from the illocutionary effect of
utterance (or by the utterance itself). And we call them "mutual" be-
hearer uptake.
cause Sand H not only both have them, they believe they both have
This and the next few chapters will be devoted largely to elaborating
them and believe the other to believe they both have them. 4 The con-
this conception of illocutionary acts as being performed with the inten-
textual beliefs that figure in speakers' intentions and hearers' inferences
tion that the hearer identify the very act being performed. In particular,
must be mutual if communication is to take place. Otherwise, it would
since the hearer's primary, but not exclusive, basis for identifying the
not be clear to each that the other is taking this belief into account. For
speaker's illocutionary intention is what the speaker says, we must
instance, if e is ambiguous and S is not punning or otherwise speaking
spell out the connection between the locutionary and the illocutionary
ambiguously, only one meaning of e will be operative; only one will be
act, such that the hearer can reasonably be expected by the speaker to
intended by S to be recognized by H as relevant. Suppose S utters "I
identify the illocutionary act being performed.
had the book stolen," intending to say that he (S) got someone to steal
the book for him-a book that S wanted to acquire. For communica-
1.2. SIMPLE VERSION OF THE SPEECH ACT SCHEMA (SAS)
tion to succeed, H must recognize that that is what S intended to say,
and not that S intended to say that S had the book stolen from him (S).
We view linguistic communication as an inferential process. The
To reasonably expect his utterance to be taken this way, S must believe
speaker provides, by what he says, a basis for the hearer to infer what
not merely that he wanted to acquire the book but also that H believes
The Theory 6 Schema for Speech Acts
this and believes that S believes this. And for H to take the utterance as
it is intended, H must believe not only that S wanted to acquire the
book but also that S believes this and believes that H believes this.
Thus, if the belief that S wanted to acquire the book is mutual, Scan
reasonably intend H to take, and H can reasonably take, S's utterance
as saying that S had someone steal the book.
In general, a mutual contextual belief figures in the speaker's inten-
tion and the hearer's inference in the following way: if p is mutually
believed between Sand H, then (1) not only do Sand H believe p, but
(2) each believes that the other takes it into account in his thinking, and
(3) each, supposing the other to take p into account, supposes the other
to take him to take it into account. Whether or not p is something
previously believed by Sand H (much less previously mutually be-
lieved), both Sand H cannot but think of p, S in making his utterance
and H in hearing it, and therefore each supposes that the other cannot
fail to take it into account and also that the other cannot fail to suppose
that he takes it into account.
The stolen book example illustrates how an MCB can be utilized by
H to close the gap between what the speaker utters and what he says.
An MCB can be utilized also to determine the intended type of illocu-
tionary act being performed. An utterance of "I love you like my
brother" might, depending on the context, have the force of an assur-
ance, an admission, an answer (to a question), or even a promise. Or it
might have merely the force of a simple assertion (and by "force" we
simply mean 'illocutionary act type'). Whichever way it is to be taken,
the speaker must intend the hearer so to take it on the basis of certain
MCBs. For example, it might be intended (and be taken) as an assur-
ance if Sand H mutually believed that H doubts that S loves him. It
would be intended and be taken as an answer if they mutually believed
that H has just asked S how he feels about H.
In short, the hearer relies on, and is intended to rely on, MCBs to
determine from the meaning of the sentence uttered what the speaker is
saying, and from that the force and content of the speaker's illocution-
ary act. Accordingly, the inference H makes and is intended to make is
of roughly the following form:
Basis
Ll. S is uttering e. hearing Sutter e
L2. S means such-and-such bye. Ll, MCBs
L3. S is saying that so-and-so. L2, MCBs
L4. S is doing such-and-such. L3, MCBs
7
Cast in this preliminary form this inference pattern constitutes what we
call the speech act schema (SAS).
In addition to mutual contextual beliefs, there are two general mutual
beliefs that the hearer relies on to make his inference. They are shared
not just between Sand H but among members of the linguistic commu-
nity at large. Pervasive as they are, they may seem almost too obvious
to mention, but must be included in the SAS. We call them the linguis-
tic presumption (LP) and the communicative presumption (CP).
Linguistic Presumption (LP): The mutual belief in the linguistic com-
munity CL that
i. the members of CL share L, and
ii. that whenever any member S utters any e in L to any other member
H, H can identify what S is saying, given that H knows the meaning(s)
of e in L and is aware of the appropriate background information.
If the LP did not prevail in CL, and between Sand H in particular, then
H could not assume that e means to S what it means to himself, and H
could not assume that S assumes he (H) assumes this. Similarly, S
could not reasonably intend to be saying that so-and-so to H in virtue of
the fact that e means so-and-so to L. Thus, in addition to the first two
lines of the SAS (L land L2), the LP is needed to license H's inference
to L3 of the SAS. To license L4 we need the CP.
Communicative Presumption (CP): The mutual belief in CL that when-
ever a member S says something in L to another member H, he is doing
so with some recognizable illocutionary intent.
If H does not think the CP is operative in a given context-if, for
instance, H thinks S is merely reciting a speech-then H has no reason
to infer any particular illocutionary intent from what S utters. The CP
does not help H determine what S's illocutionary intent is-H must
rely on what S says and on the MCBs for that. The CP licenses only H's
conclusion that S has some illocutionary intent or other. Accordingly,
we may augment our provisional version of the SAS. S intends H to
reason as follows:
Basis
Ll. S is uttering e. hearing Sutter e
L2. S means such-and-such bye. Ll, LP, MCBs
L3. S is saying to H that L2, LP, MCBs
so-and-so.
L4. S is doing such-and-such. L3, CP, MCBs
The Theory 8 Schema for Speech Acts
In the following section we raise certain issues pertaining to the
connection between what S says (L3) and what S means (L2) by what
he utters (L1).5 The relation between the locutionary act (L3) and the
illocutionary act (L4) is discussed in section 1.4.
1.3. SAYING AND THE LINGUISTIC PRESUMPTION
Without mutually believing that they share the language they are using,
people would not, and perhaps could not, use the language to commu-
nicate: the third step in the SAS (L3) would be blocked. Generally, this
mutual belief between Sand H arises from the linguistic presumption
that prevails among members of the community at large. As a matter of
social fact, the LP in a community is so strong that not to know the
language is often a sign of nonmembership in the community. People
presume that if you belong to the community, you know the language.
So when S utters something e in L (the language in question), he
expects H to understand it. Indeed he expects this not because he
thinks H has heard e before or ever learned the meaning of e in par-
ticular, but because he thinks H knows L and will, by virtue of knowing
L, understand e. 6 Thus, because the LP applies generally to communi-
cation situations in CL, Sand H mutually believe that each will under-
stand almost anything in L uttered by the other; unless something
happens to show that the LP does not apply, Sand H are each in a
position to reach L3 of the SAS, H to identify what S is saying and S to
intend H to identify what he is saying.
Implicit in our discussion of the LP is the distinction between a
group's having a language and their sharing that language. However
improbable, a group of people could all have a language without mutu-
ally believing they do, in which case they probably would not use the
language to communicate- no one would have any reason to believe
he would be understood. Because of the distinction between having a
language and sharing it, we cannot expect linguistic meaning to be
explicated in social terms. It is logically possible for a person to know a
language without ever having used it (or heard it used) to perform
speech acts. 7 Although the concepts of a language, of knowing a lan-
guage, and of sharing a language figure in the linguistic presumption
(the mutual belief prevailing in a community that a certain language is
known and shared), we need not ascribe theoretical understanding of
these concepts to ordinary speakers, who, after all, are not philos-
ophers or linguists. Intuitive understanding is enough for them. Though
9
not pertinent here, characterizing what a language is and what it is
to know one (linguistic competence) presents tough philosophical
problems.
Recall that in the speech act schema, S's uttering a sentence e' and
the meaning of e are covered by L1 and L2, while what S says is l~ft for
L3. L2 is separate from L1 for two reasons: that S means anything 'at all
bye-that S is doing anything over and above the act of uttering-
requires further intentions; moreover, what S means by e may not be
wholly determined by the semantics of e, since e may be ambiguous.
So whereas S's act of uttering e is reported by direct quotation in L1 of
the SAS, the operative meaning of e is given in L2. Because linguistic
meaning does not in general determine reference, S's locutionary act is
represented separately by L3, in the form of indirect quotation: S said
that so-and-so. For this, references must be specified. The pattern of
inference whereby H identifies what S says will be spelled out in chap-
ter 2.
As familiar as indirect quotation is, so are the problems it gives rise
to. 8 Nevertheless, we will provisionally assume that the notion of in-
direct quotation can be made philosophically acceptable enough to be
used in the SAS. For present purposes let what S means by e be
represented by a lacuna of undetermined form, " ... ," and let what S
says in uttering e be represented by a dummy indicator for sentence
type "*" and "p" for a proposition: "that *( ...p ... )." This notation is
meant to indicate that what is said is a function of the intended meaning
(" ... ") of the expression e. If e is declarative, then what is said may be
specified by truth conditions. For other sentence types it may be feasi-
ble to generalize the notion of a truth condition and thereby allow a
homogeneous semantics for natural languages and a single style of
specification for what is said (see Stenius 1967, Lewis 1969, Straws on
1971, and Katz 1972). If e is imperative, the that-clause specifying what
S says (that !( ..p ... )) is not itself imperative but of the form, "that H is
to A."
As for interrogative sentences, there seem to be two options. On the
one hand it could be argued that sentences like "What time is it?" do
not express a proposition and their use is to be reported with "ask"-S
asked (H) what time it is. In this view (Schiffer 1972, 114ft) interroga-
tive sentences are conventional means for performing illocutionary acts
of just one particular kind, namely, asking a question. If this is correct,
then the locutionary step in the SAS (" S said that ... ") is simply by-
passed in the case of interrogatives, and indirect quotation will be of
The Theory 10
the form" S asked .... " Nevertheless, such sentences need not be used
literally (as when used rhetorically to make a statement) and so an
adaptation of the SAS must allow for that. On the other hand it might
be suggested (as by Katz 1977c, 205ff.) that what is said when S uses an
interrogative expression like "What time is it?" is: that H is to tell S
what time it is. In general, the form of the report will be: S says that H
is to tell S - - , where the blank is filled in by some expression deter-
mined bye. 9 On this account questions would be a particular case of
requests and would be performed normally via the schema. Since either
account is compatible with our overall theory, we will leave the matter
as it is for now.
1.4. LITERAL ILLOCUTIONARY ACTS
In the speech act schema L3 represents what the speaker says and L4
what he is thereby doing in saying it. Since the speaker might not be
performing any illocutionary act at all, it is only on the presumption
that he is (the CP) that the hearer will infer that the speaker is perform-
ing some illocutionary act or other. As for identifying what the act is,
the hearer relies primarily on what is said, and we find the most
straightforward relation between what is said and what is done when
the speaker means what he says and nothing else. In this case he is
speaking literally and what he does is largely determined by w~at he
says.
Because of nonliteral and indirect illocutionary acts, the slogan
"Meaning determines force" is generally false. It is most nearly correct
in the case of literal acts-but not quite. Although what the speaker
does might be determined by what he says, that he is performing any
illocutionary act at all is not; he could be merely practicing his English
or mechanically reciting some lines. Moreover, that he is speaking
literally is not determined by what he says. If S says, for example,
"The sun is shining on me today," he could be talking either about the
weather or about his fortunes, depending on whether or not he is
speaking literally. Inasmuch as he can use the same sentence literally or
nonliterally, how he intends his utterance to be taken is not determined
by what he says.
Even allowing for the fact that the meaning of what is uttered does
not determine that some illocutionary act is being performed, much less
that it is being performed literally, it is not always true that meaning
determines the force of literal illocutionary acts. In general, the mean-
ing merely delimits the force. For example, if someone says that he will
Schema for Speech Acts 11
return, whether he is making a promise or merely a statement of inten-
tion, his illocutionary act is literal. So the force (illocutionary act type)
of an utterance need not be explicit to be literal. You do not have to say
"I accuse ... " to make an accusation. For that matter, you can use a
performative verb nonliterally, as when posing a threat by saying "I
promise."
Let us borrow Searle's (1969, 31) notation for representing an illocu-
tionary act by "F(P)," where "F" represents the force and "P" the
propositional content of the illocutionary act (lowercase "p" repre-
sents the proposition in the locutionary act). Now suppose that S utters
e, which means' .. .', and thereby says that *( .. .p .. .). His act is literal,
and represented by "F*(p)," 10 just in case the proposition that P is
the same as the proposition that p and the illocutionary force F of the
utterance is locutionary-compatible (L-compatible) with the sentence
type and meaning of e. Without giving a definition, we can introduce
the notion of L-compatibility by examples. An utterance's being a pre-
diction is L-compatible with the sentence used only if the sentence
contains future time reference. If that sentence contains an action verb
predicated of S, then the force of the utterance is L-compatible with the
sentence whether the sentence is used to make a promise or a predic-
tion. ll An utterance's being a request or an order is L-compatible only
with imperative sentences; analogously, an utterance's being a ques-
tion is L-compatible only with interrogative sentences. If an utterance
has a force L-incompatible with the mood and meaning of the sentence
used, it is not literal. Notice that our characterization of literal utter-
ances requires what the speaker says to be the same as what he F's
that is, that p be the same proposition as P. Even though an utteranc~
of the sentence, "I am sorry for stepping on your toes," has the
L-compatible force of an apology, the speaker can be apologizing non-
literally (say, for preempting the hearer's authority).
U sing our notation we can reformulate our provisional version of the
SAS. In so doing we will recast L4 in such a way that it is left open for
further inference, whether or not S is speaking literally. S intends H to
reason as follows:
Basis
Ll. S is uttering e. hearing Sutter e
L2. S means ... bye. L1, LP, MCBs
L3. S is saying that *( . ..p ... ). L2, LP, MCBs
L4. S, if speaking literally, is
F*-ing that p. L3, CP, MCBs
The Theory
12
Only after we spell out the details of these steps and present our
taxonomy of illocutionary acts will we be in a position to elaborate the
SAS to include the further steps whereby H infers, as S intends him to
infer, what illocutionary act is in fact being performed, be it literal or
otherwise. As we will see, H relies on the presumption of literalness
(PL):
Presumption of Literalness (PL): The mutual belief in the linguistic
community CL that whenever any member S utters any e in L to any
other member H, if S could (under the circumstances) be speaking
literally, then S is speaking literally.
If it is evident to H that S could not be speaking literally, H suplJoses S '
to be speaking nonliterally and seeks to identify what that nonliteral
illocutionary act is.
1.5. THE COMMUNICATIVE PRESUMPTION AND
ILLOCUTIONARY INTENTIONS
The communicative presumption is the mutual belief prevailing in a
linguistic community to the effect that whenever someone says some-
thing to somebody, he intends to be performing some identifiable illo-
cutionary act. We say "to the effect that" because, of course, people
don't have the technical concept of illocutionary acts and therefore do
not have beliefs, much less mutual beliefs, about illocutionary acts. But
they do mutually believe that speakers speak with overt intentions, and
this mutual belief figures in ordinary communication situations. People
do rely on others to have identifiable intentions in their utterances, and
they expect others to rely on them to have such intentions.
There are all sorts of effects a speaker can intend an utterance to
have on the hearer. S mayor may not intend H to recognize S's
intention to produce a certain effect, and even if he does so intend, H's
recognition of S' s intention may be incidental to the production of that
effect. In general, hearer recognition of perIocutionary intentions is
incidental to the production of perIocutionary effects. Even in the spe-
cial case where identification of the speaker's intention is necessary to
the production of a perIocutionary effect-H might believe something
or do something because and only because S wants him to-still there
is a distinction between the hearer's recognizing that intention and its
being fulfilled. The hearer might recognize what effect is intended
without its being produced in him. What distinguishes iIlocutionary
Schema for Speech Acts 13
intentions, we suggest, is that their fulfillment consists in their recog-
nition.
This general conception of illocutionary acts and intentions is shared
by Searle and Strawson (in detail their views differ radically from ours).
Searle (1969, 47) points out the connection between the fulfillment of
illocutionary intentions and their recognition when, in contrasting illo-
cutionary with perlocutionary acts, he says,
In the case of illocutionary acts we succeed in doing what we are trying
to do by getting our audience to recognize what we are.trying.to d~. But
the' effect' on the hearer is not a belief or a response, It CO~SISt~ sImply
in the hearer understanding the utterance of the speaker. It IS thIS effect
that I have been calling the illocutionary effect.
As Strawson puts it (1964, 459),
The understanding of the force of an utteran~e in al! cases .involyes
recognizing what may be called broadl~ an audIence-dIrected. mtentlon
and recognizing it as wholly overt, as mtended to be recogmzed.
Their formulations help to spell out Austin's view that successful com-
munication in performing an illocutionary act consists in uptake, that
is, in the hearer identifying the illocutionary act being performed. Our
later elaboration of the SAS will detail the pattern of inference by
which this is accomplished, but first we must consider precisely what
sort of intention is such that its fulfillment consists in its recognition.
What sort of intention is distinctively illocutionary and communica-
tive?
Both Searle and Strawson suggest that this intention is essentially
reflexive and of the sort discovered by Grice (1957). According to
Searle (1969, 43),
In speaking I attempt to communicate certain t~ings .to my hear~r by
getting him to recognize my intention to communIc~te J~st those thm~s.
I achieve the intended effect on the hearer by gettmg hIm to recogmze
my intention to achieve that effect.
Not just any way of achieving that effect will do. Hypnosis or electrical
stimulation might "get" the hearer to recognize the speaker's intention,
but for this recognition to be the effect of linguistic communication, it
must be achieved by an inference from the speaker's utterance, and
normally that is how it is accomplished.
Searle criticizes Grice's account of speaker meaning in terms of re-
flexive intention, that is, in terms of the intention "to produce some
The Theory 14
effect in an audience by means of the recognition of this intention"
(Grice 1957,385). Searle argues that the sorts of effects Grice mentions,
such as beliefs, intentions, and actions, are not produced by means of
recognition of the intention to produce them. For example, the hearer
might recognize that he is to believe something and yet refuse. These
sorts of effects are perlocutionary, and the speaker's illocutionary act,
whose identity he is trying to communicate, can succeed without the
intended perlocutionary effect (if there is one) being produced. So a
reflexive intention is involved in communication, just as Grice claimed,
but the kinds of intended effects he specified are not of the right sort.
Getting the hearer to recognize them does not constitute producing
them. In section 1.6 we consider just what sort of reflexive intention is
fulfilled merely by being recognized.
Grice's account of reflexive intentions in communication neglects the
role of the communicative presumption when the communication is
linguistic. 12 Grice focuses on nonlinguistic examples like drawing a
picture and deliberately frowning. 13 When people do things like these,
there is no presumption that they have a communicative intention, as
there is in the case of linguistic utterances. Because of the CP, when
somebody says something to someone, he cannot but expect-he need
not intend- the hearer to think he has some identifiable illocutionary
intention. Contrary to Grice's nonlinguistic cases, H's reason for
thinking S has some such intention is not that he has spotted anything
special in S's utterance but, because of the CP, merely that S has
uttered something linguistic. S realizes that H routinely assumes that
some recognizable intention is there, so no generic intention to be
performing some such act is necessary. Indeed, if S were mimicking
someone or rehearsing a line and thought this was not evident to H
(with the implicit understanding that the CP was inoperative), S would
have to have a special intention, one that he could reasonably expect H
to recognize, not to be performing a full-blown illocutionary act. Mter
all, being a presumption, the CP is operative unless there is indication
to the contrary (as ordinarily there is when people mimic or recite).
The difference between linguistic communication and Grice's non-
linguistic cases is the presence of a presumption that there is an in-
ference to be drawn as to what the speaker is doing in issuing his
utterance. In Grice's examples part of what the audience has to infer is
that there is an inference to be drawn. The SAS, which includes the CP,
represents the nature of this inference for linguistic cases, and the
hearer implements this inference by recognizing what the speaker ut-
Schema for Speech Acts 15
ters. That it is a sentence in a shared language is enough to implement
the inference. So although reflexive intentions (R-intentions) are es-
sential to linguistic communication, not just any sort ofR-intention will
do. Linguistic R-intentions are executed pursuant to the communica-
tive presumption, and their fulfillment consists in their recognition. We
must now consider what their content can be for this to be true.
1.6. ILLOCUTIONARY INTENTIONS AND EFFECTS
An illocutionary act is communicatively successful if the speaker's i110-
cutionary intention is recognized by the hearer. These intentions are
essentially communicative because the fulfillment of illocutionary in-
tentions consists in hearer understanding. Not only are such intentions
reflexive, their fulfillment consists in their recognition. Thus the in-
tended effect of an act of communication is not just any effect produced
by means of recognition of the intention to produce a certain effect, it is
the recognition of that effect. There seems to be a reflexive paradox
here, but in fact there is none. The effect, the hearer's recognizing the
speaker's intention to produce that effect, is not produced by the
hearer's recognizing that intention-that would be worse than a para-
dox, it would be a miracle. Rather, it is produced by the hearer's
recognizing that the speaker has an intention to produce a certain effect
in him that he is to identify (and thereby have produced in him) partly
by recognizing S's intention to produce an identifiable effect. The
hearer has to figure out what that intention-the intended effect-is,
on the basis primarily of the speaker's utterance, along the lines of the
SAS.
Now what sorts of (intended) illocutionary effects-effects consist-
ing in recognition ofR-intentions-can there be? In other words, what
can be the content of communicative intentions? It is a commonplace
that linguistic communication consists in putting one's thoughts into
words. This cliche is correct as far as it goes; the problem is to go
further. In our view, to communicate is indeed to express a thought or,
more generally, an attitude, be it a belief, an intention, a desire, or even
a feeling; but in saying that to communicate is to express an attitude,
we mean something very specific by "express."
Expressing: For S to express an attitude is for S to R-intend the hearer
to take S's utterance as reason to think S has that attitude.
The Theory 16 Schema for Speech Acts 17

Accordingly, the intended illocutionary effect (or simply illocutionary the design, intention, or purpose of producing them . . . We shall call
intent) is for H to recognize that R-intention. In the taxonomy of com- the performance of an act of this kind the performance of a perlocu-
municative acts that we develop in chapter 3 many types of illocution- tionary act.
ary acts are differentiated by types of attitudes expressed.
Since there is virtually no limit to the sorts of things that can result from
For now, consider a couple of simple and common types of illocu-
speech acts-almost anything is possible, from in~u1ting someone to
tionary acts: statements and requests. In the case of statements the
starting a war-it would seem reasonable to restnct the category of
speaker expresses two attitudes: belief in a certain proposition and the
perlocutionary acts in whatever ways seem theoretically appropriate.
intention that the hearer believe it as well. That is to say, for S's
This is a matter of terminological stipulation, of course, but that does
utterance of e to be a statement that P, S must R-intend H to take the
not make it arbitrary.
utterance as reason to think (a) that S believes that P and (b) that S
We propose first to limit perlocutionary acts to the intentional pro-
intends H to believe that P. Correlatively, for H to understand that S is
duction of effects on (or in) the hearer. Our reason is that only refer-
stating that P in uttering e, H must take S's utterance of e as R-
ence to intended effects is necessary to explain the overall rationale of a
intended to be reason to think (a) and (b). For a statement to have been
given speech act. Utterance, locutionary, and illocutionary acts are all
made and to be successful as an act of communication, it is not neces-
intentional and are generally performed with the primary intention of
sary that H actually think that S believes that P or that H believe that P
achieving some perlocutionary effect. To be sure, a speaker can insult,
himself. These would be perlocutionary effects of S' s utterance and are
appease, disturb, or excite someone without intending to, but unless
not necessary for the success of the illocutionary act of stating. We
this is done intentionally, the fact that it is done does not help explain
might say, speaking loosely, that S's statement was unsuccessful (with
the speech act. In any case, the vocabulary of verbs of speech actions
respect to a furtherperlocutionary effect) unless H believed that P
cannot be relied on to mark the distinction between illocutionary and
(presumably taking S to believe that P); but surely it would be correct
perlocutionary acts.
to claim that S had successfully made a statement if H understood S's
We propose further to restrict perlocutionary acts to producing ef-
utterance of e, even if H didn't believe that P. It is sufficient that H
fects from steps of the speech act schema. In our preliminary version of
recognize S's R-intention, S's expressed attitudes. This is what com-
the SAS these steps include the hearer's identification of L1, the utter-
munication is about; anything more is more than just communication.
ance of e; L2, what S meant bye; L3, the locutionary act; and L4, the
Similarly, for S' s utterance to count as a request that H do A, S must
illocutionary act. We will be interested primarily in perlocutionary ef-
R-intend H to take S's utterance as reason to think (a) that S desires H
fects generated from L4, effects that rely on hearer uptake (of course
to do A and (b) that S intends H to do A because of S's desire. His
their production involves more than uptake, more than recognition of
request is successful as an act of communication if H recognizes S's
the intention to produce them). Perlocutionary effects can be generated
R-intention. Again, anything more-H's actually doing A-is more
from other steps of the SAS as well. The utterance of certain words
than just communication.
might be intended to offend someone just by their sound or their man-
ner of pronunciation; or perhaps their meaning is what offends. And the
1.7. PERLOCUTIONARY ACTS AND EFFECTS
locutionary act might have a distinctive perlocutionary effect, such as
reminding the hearer of a person or event referred to. When we refine
Austin (1962, 101) introduces the notion of a perlocutionary act as
the SAS, we will see that there are other ways, corresponding to steps
follows:
in the schema as elaborated, in which perlocutionary effects can be
Saying something will often, or even normally, prod,uce certain co~ generated. For example, the very fact that an illocutionary act is per-
sequential effects upon the feelings, thoughts, or actIons of the audI- formed nonliterally or indirectly might have a definite perlocutionary
ence, or of the speaker, or of other persons: and it may be done with effect, such as protecting the hearer's feelings or making him sus-
picious.
The Theory 18

This second restriction, construing as perlocutionary only those in- Chapter Two Elaborating the Schema:
tended effects generated off of steps of the SAS, excludes all sorts of Locutionary Acts
other speech acts, which we will survey in section 5.5: joking, manip-
ulating, boring, interrupting, and so on. Some of these speech acts are
essentially intentional, some even R-intentional. However, they do not
work off of the SAS, and in some cases, such as in telling a joke, they
presuppose the suspension of the communicative presumption.
Our characterization of the various aspects of speech acts as well
as our formulation of the speech act schema itself should be taken as
provisional. In the next chapter we spell out the details of each step as
formulated thus far, thereby enabling us to refine our conception of
locutionary acts. In chapter 3 we develop a taxonomy of communica-
The locutionary act, the act of saying something, provides the hearer
tive illocutionary acts, followed in chapter 4 by our full formulation of
with the core of information from which to infer the speaker's illocu-
the SAS to cover nonliteral and indirect as well as literal illocutionary
tionary (communicative) intent. Other items of information contribute
acts. In chapter 5, after the SAS is fully elaborated, we explain how it
substantially to this identification, especially when S is speaking non-
contributes to a philosophical analysis of linguistic communication and
literally or indirectly. But even when he is speaking literally, such that
provides a framework for the psychological explanation of linguistic
his illocutionary intent is made more or less explicit by what he says,
communication. The SAS is only a schema, however, and can do only
his intent still has to be inferred by the hearer. Thus, a locutionary act is
so much. It represents the pattern of inference made by the hearer but
always distinct from any literal illocutionary act being performed, and
it does not represent how the inference is made. In particular, even
until the hearer takes into account other information besides that pro-
though mutual contextual beliefs are cited in various lines of the
vided by the locutionary act, all he can infer is what, if any, literal
schema and are relied on by the hearer to go from one step to the next,
illocutionary act is being performed, as indicated in L4 of the SAS.
the SAS is not equipped to predict which MCBs are activated and so
cannot predict precisely how a given hearer will take a given utterance.
2.1. AUSTIN ON LOCUTIONARY ACTS
Moreover, although it represents the pattern of inference in steps,
thereby organizing the mass of information available to the hearer, in
Austin distinguishes three aspects of the locutionary act.
practice the hearer often works holistically, both looking ahead and
backtracking as he goes along. Our examples often illustrate this, but To say anything is:
we will not return to the status of the SAS as a whole until it is spelled (A.a) always to perform the act of uttering certain noises (a 'phonetic'
out in detail. act), and the utterance is a phone;
(A.b) always to perform the act of uttering certain vocables or word~,
i.e. noises of certain types belonging to and as belonging to a certam
vocabulary, in a certain grammar, with a certain intonation, &c. This
act we may call a 'phatic' act; and
(A.c) generally to perform the act of using that [sentence] or its con-
stituents with a certain more or less definite 'sense' and a more or less
definite 'reference' (which together are equivalent to 'meaning'). This
act we may call a 'rhetic' act. (1962, 92-93)
Unfortunately, there are two ways of taking the phrase "with a certain
sense and reference" and thus two ways of taking the notion of a rhetic
(hence of a locutionary) act. On one reading, the phrase identifies the
The Theory 20 Locutionary Acts 21

operative sense (and denotation) of expressions in case they are ambig-
uous; on the other it specifies what the speaker means and refers to by
the expressions used. The latter is probably what Austin had in mind
(pp. 114-115, note 1). Even though he did not draw the distinction
clearly, for Austin there is a major break betwe'en the phatic act and the
rhetic act, in that specification of the former entails no specification of
what the speaker meant, whereas the latter does. For one thing, iden-
tifying the operative sense (and attendant denotations) of the expres-
sions in a sentence uttered does not guarantee that something was said
rather than, say, recited. Moreover, because few referring expressions
uniquely pick out particular referents solely in virtue of their meaning,
reference at the level of the phatic act does not in general determine
reference at the level of the rhetic act. Though Austin did not draw
his phatic/rhetic distinction quite right, he had the right ingredients:
the (operative) meaning of ambiguous expressions, what the speaker
meant, and the fixing of referents. But how should these be blended
into a more adequate formulation?
We will attempt to answer this question by spelling out how the
hearer reaches line L3 of the SAS, the step at which he identifies what
the speaker says. As provisionally formulated, the SAS runs as follows.
S intends H to reason:
Basis
Ll. S is uttering e. hearing Sutter e
L2. S means ... bye. Ll, LP, MCBs
L3. S is saying that *( ...p ... ). L2, LP, MCBs
L4. S, if speaking literally, is
F*-ing that p. L3, CP, MCBs
We will examine the details of the inference from one line to the next.
As we proceed, it should be understood that the speaker's intention
cannot realistically be supposed to include every such detail.
2.2. INFERRING OPERATIVE MEANING
Hearer H, at Ll of the SAS, realizes that S has uttered e. 1 Assuming S
meant anything at all bye, to reach L2 H must determine what S meant
bye, that is, the operative meaning of e. Since the linguistic presump-
tion is in effect, H can reasonably suppose that S did mean something
by e and that this is something that e means in L, their shared language.
However, inasmuch as ambiguity is rampant in natural languages, H is
likely to need more than the LP and the identity of e to determine what
S meant bye. If e has two meanings in L, H has to make an inference of
the following form (how the lacunae are filled- the vexing problem of
representing linguistic meaning-will be discussed in section 8.1):
Basis
Ll. S is uttering e. hearing Sutter e
(a) e means ... and __ in L. knowledge of L
L2. S means ... bye. Lla, LP, ?
Only if e is unambiguous can H infer L2 on the basis of what e means in
L, and even then he must assume, pursuant to the LP, that Stakes e to
mean ... in L. Disambiguation is not required in the unusual case where
e is ambiguous but S intends both its meanings to be operative. Even
then H needs a basis for inferring that both of e's meanings are indeed
operative.
Consider the case, by far the most common, where e is ambiguous
but only one sense is operative. How does H infer which meaning is
operative? It would seem that he must reject all but one of the meanings
of e as contextually inappropriate and rely on certain mutual contextual
beliefs to do this. Accordingly, his inference would take the following
form:
Basis
Ll. S is uttering e. hearing Sutter e
(a) e means ... and __ in L. Ll, knowledge of L
(b) S means ... bye, or S
means __ bye. Lla, LP
(c) The supposition that S
means __ by e is
contextually less appropriate. LIb, MCBs
L2. S means '" bye. LIb, Llc
We do not claim that contextual selection always represents a psycho-
logically real process (see chapter 11), nor, where real, that selection
necessarily proceeds in just the sequence specified. Usually we seem
just to hear (or read) and understand the expression e in the contextu-
ally most appropriate way-which is why we often miss subtle puns.
Perhaps, then, the process of understanding an utterance involves
operations that make certain readings more probable, given certain
mutual contextual beliefs, and as hearers we often take these readings
as first hypotheses concerning what S meant by e unless (or until) they
are defeated by MCBs, future remarks, and so on. For instance, sup-
The Theory 22
pose we are discussing a local airport's flight pattern across the univer-
sity campus and I say "Flying planes sure can be dangerous." Given
the supposition that I am being relevant, one of the things I might have
meant bye is ruled out-discussing the hazards of piloting would have
been irrelevant and would have changed the course of conversation in
an illegitimate way (see chapter 4). This leaves the other reading as
more plausibly operative. So what we are representing here is the
information available to H and its general direction of flow, not the
operations underlying H's use of the information. But what would be
the structure of H's inference if he made it step by step?
To reach L 1(a), H simply relies on his knowledge of L. To reach
Ll(b) he relies on the LP and supposes that Stakes e to have just those
meanings that in fact it has in L. Getting to Ll(c), at which H must
exclude certain meaning(s) of e as inappropriate, is more problematic.
Of course, if e is unambiguous, this step is vacuous, and if S seems to
mean both ... and __ bye, this step is bypassed. But in the case of
ambiguous expressions uttered with a single operative meaning, H
must select one of the disjuncts from Ll(b) as the operative one. As the
flying planes example illustrates, under the circumstances of the utter-
ance, H may have certain expectations or make certain judgments of
appropriateness to decide which meaning of e is operative. If H' s con-
ception of the conversational situation is sufficiently determinate or his
expectations of the direction of S's utterance sufficiently specific, H
may not even go through the process of considering and discarding the
inoperative meanings. In any case, if H's antecedent expectations do
not settle the matter, then contextual appropriateness must, or H can-
not but infer that S is speaking ambiguously. (We are being vague here
about what contextual appropriateness is; this notion will be examined
at length in our discussion of the illocutionary part of the SAS in chap-
ter 4.)
As the intermediate steps from Ll to L2 are represented in the SAS,
H takes for granted that whatever S means by e is one of the meanings
of e in L. Or, rather, at Ll(c) he does not reject all the meanings of e in
L for being contextually inappropriate. Before concluding that the lin-
guistic presumption is inoperative (very unlikely if the utterance occurs
in the middle of a conversation), H will ask himself (in effect), "Does S
think e means something different from what it means in L, or is he
simply speaking nonliterally?" He has followed the SAS to the con-
sequences of literality and, having ruled them out, searches for a
nonliteral interpretation of S's utterance. If he can find no such in-
Locutionary Acts 23
terpretation, he may backtrack to Ll(b) and consider the possibility of
S's being mistaken about the meanings of e in L. To illustrate how
complicated these matters can get, imagine a case like the following, in
which S utters "The vote was anonymous." H has four distinct op-
tions:
(1) a. S thinks e has the meaning 'It was a secret ballot' in English,
S was speaking literally, and this is what he was saying (literal
meaning).
b. S was speaking sarcastically, not literally, and was nonliterally
stating that it was a public vote (nonliteral meaning).
c. S (mistakenly) thinks "anonymous" means 'unanimous,' and, if
speaking literally, was saying that everyone voted for it (false lin-
guistic belief).
d. S meant to utter "unanimous" and to say that everyone voted for
it (slip of the tongue).
Such lines of inference can converge, in which case it will not matter
which route the hearer has taken. In this example it does not matter
which of (1c) or (1d) was followed; the result is the same.
Since we have been concerned with the inference from what the
speaker utters to what he means by it under the circumstances, we have
looked at the meanings of e from H's point of view. In the example, H's
option (1c) involves attributing to S a false belief about the meaning of
what he uttered. The reverse can occur too, as is evident if we take S's
point of view. Suppose S believes that H takes e to mean ... but does
not himself believe that. S can still communicate linguistically by
playing along with H's linguistic belief, which he (S) takes to be mis-
taken. If S is correct about H's belief about e, H won't know the
difference and will infer the operative meaning of e even though e has
no such meaning in L. Finally, Sand H can mistakenly, though mutu-
ally, believe that e means ... in L. Perhaps S is the parent of H so that
H acquired S's linguistic misconceptions. In such a case neither S nor
H would know the difference: S would mean ... by e even though e
does not mean ... in L; H would take e to mean ... in L and infer that S
means ... bye; and nothing would occur to suspend the linguistic
presumption. 2 These various cases in which there is a discrepancy
between the operative meaning of e and what e means in L seem to be
cases of genuine communication. This being understood, as presented
so far the SAS represents only the normal case where what S means by
e is one of the meanings of e in L.
The Theory 24
2.3. INFERRING LOCUTIONARY ACTS
Hearer H, at L2 of the SAS, has inferred that S means ... bye. From
this together with the LP and various MCBs, H must infer that S is
saying that *( ...p ... ):
L2. S means ... bye.
L3. S is saying that *( ...p ... ). L2, LP, MCBs
At L3, H is able to determine that something is being said and to
identify what is being said. Since the LP sanctions the first part of this
inference, our problem is to show how the second part works. That is,
what is the pattern of inference by which H identifies what S is saying?
To identify what is said is to identify sentence mood and proposi-
tional content. Three things determine propositional c<?ntent: operative
meaning(s), referent(s), and time(s) specified. On this initial account, to
identify what is said would be to identify (in addition to sentence mood)
what referent(s) are being ascribed what properties (or relations) at
what time(s). It is clear that H's knowing the language L and believing
that S means ... by e suffices to fix:
1. which mood the sentence is (I-, ?, !),
2. which parts of e can be used to refer (the referring expressions in e),
3. which parts specify times,
4. the senses of the referring expressions,
5. the properties and relations being ascribed.
Since these are determined by what S means bye, all that is left for H
to identify in order to determine what S said is the type of saying
(corresponding to the sentence mood), the objects being referred to, and
the times being specified. With regard to times-except for tenses,
whose contribution to what is said H can determine relative to the
occasion of utterance-temporal descriptions, pronouns, and adverbs
can be subsumed for our purposes under the heading of referring ex-
pressions. Accordingly, given that the items listed are fixed by what H
takes to be the operative meaning of e, what remains for H to identify
in order to determine what S said are sentence mood and references.
2.3.1. Sentence Mood and Type of Saying
Since sentence mood contributes in a regular way to what is said, we
Locutionary Acts 25
adopt the following form for representing this contribution:
Declarative e:
S is saying that I-( ...p ... );
S is saying that it is the case that ( ...p .. .).
Imperative e:
S is saying that !( ...p ... );
S is saying that H is to make it the case that ( ...p ... ).
Yes/No Interrogative e:
S is saying that ?( ...p ... );
S is asking (or saying that H is to tell S) whether or not it is the case that
( ...p ... ).
Wh Interrogative e:
S is saying that ?( ... Wh-x p ... );
S is asking (or saying that H is to tell S) ( ... Wh-x p .. . ).
For Wh interrogatives, the symbol Wh-x next to p schematizes the
identity of the x being questioned. Thus "Who" in "Who discovered
the calculus?" becomes: the identity of the person x such that x(dis-
covered the calculus). Putting this all together yields the following
locutionary specification: in uttering "Who discovered the calculus?" S
is asking (or saying that H is to tell S) the identity of the person x such
that x discovered the calculus.
2.3.2. Reference
To identify the referent(s) is to identify what S intends to be referring
to. But how does H use the presumptions and MCBs to recognize S's
referential intent? The main devices S uses to signal his referential
intent are descriptions, pronouns, and proper names, but in general
their linguistic meaning does not determine their reference. For each
type of referring expression we describe the pattern of inference H
makes to identify the referent.
Definite Descriptions
Suppose S utters something of the form: "The so-and-so is ... "3 H
will be expected to search his memory or the context for some relevant
referent that is so-and-so and that can reasonably be thought to be what
S intends to pick out. In terms of the SAS:
The Theory 26
Ll. S is uttering "The so-and-so is ... "
L2. (a) S means 'some definite so-and-so is ... ' bye.
(b) "The so-and-so" is a definite referring expression in English.
knowledge of L
(c) In uttering e, S intends to pick out some definite x.
L2(b), LP
(d) In uttering e, S intends to pick out some definite x that is
so-and-so.
L2(a), L2(c), PL
(e) The so-and-so S intends to pick out is the such-and-such. 4
L2(d), MCBs
(f) In uttering e, S is saying of the such-and-such, under the de-
scription of being so-and-so, that it is ...
L2(e), LP
L3. In uttering e, S is saying that the so-and-so (namely, the such-and-
such) is ...
L2(f), LP
The schema now allows for a variety of referential possibilities for
descriptions. Line L2(a) allows for different uses of the definite article
as contrasted in:
(2) a. The asteroids are small planets. (specific)
b. The children are asleep. (specific)
(3) a. The Turks invaded Vienna. (nonspecific)
b. The Chinese have known about porcelain for three thousand
years. (nonspecific)
(4) a. The owl is nocturnal. (generic)
b. The kiwi is extinct. (generic)
In (2a) "the asteroids" is used to refer to the totality of asteroids,
whereas in (2b) "the children" is used to refer to a certain group of
children (determinable in the context). In both cases it is clear which
particular entities are being referred to and which are not. In (3a)
context does not help to determine which Turks "the Turks" is used to
refer to. An analogous point applies to (3b), and matters are further
complicated, in ways we will not take up here, by the fact that no
individual Chinese has known anything for three thousand years. Fi-
nally, in (4a) a certain type of animal is being referred to, and as (4b)
Locutionary Acts 27
makes clear, referring to a type is not equivalent to referring to all the
individuals of that type.
Lines L2( d) - L2(f) allow for nonliteral as well as literal uses of de-
scriptions. s For instance, S might observe that the last (adult) guests at
a party have passed out in the backyard after making fools of them-
selves and say to H "The children are asleep," intending H to infer that
S is saying that the remaining guests are asleep:
Ll. S is uttering "The children are asleep."
L2. (a) S means 'The children are sleeping' by "The children are
asleep."
(b) 'The children' is a definite referring expression of English.
(c) In uttering "The children are asleep" S intends to pick out
some definite x' s.
(d) There are no children (so-and-sos) for S to be referring to.
(e) Since the guests were acting like children, the x's S intends to
pick out are the remaining guests (the such-and-suches).
(f) In uttering "The children are asleep," S is saying of the re-
maining guests, under the description of being children, that they
are asleep.
L3. In uttering "The children are asleep," S is saying that the children
(namely, the remaining guests) are asleep.
S is not saying that the remaining gue~ts are asleep but saying of the
remaining guests, referred to as children, that they are asleep.
Pronouns
We deal here only with personal pronouns in their simplest uses. A
comprehensive account would cover demonstrative and relative pro-
nouns, and such phenomena as ostensive and cross reference. Suppose
S utters something of the form: "She is ... " As with descriptions, H
will be expected to search his memory or the context for the relevant
referent. In terms of the schema:
Ll. S is uttering" She is ... "
L2. (a) S means 'some female is ... ' by "She is ... "
(b) "She" is a singular definite referring expression in English.
(c) In uttering e, S intends to pick out some definite x.
(d) In uttering e, S intends to pick out some definite female.
(e) In uttering e, S intends to pick out the female who is so-and-so.
L3. In uttering e, S is saying that some definite female (namely, the
so-and-so) is ...
The Theory 28
As with descriptions, the schema accommodates various uses of the
pronoun, for example, "she" used to refer to inanimate objects such as
boats or, nonliterally, to males. By providing parameters for speaker
and hearer, the schema is able to represent some uses of the pronouns
"I" and "You":
(5) a. In uttering "I like you," S is saying (to H) that S likes H.
b. In uttering "Leave the room!" S is saying (to H) that H is to
leave the room.
c. In uttering "What time is it?" S is saying (to H) that H is to tell S
what time it is.
Proper Names
Suppose H hears an expression of the form: "Sam is ... " H will be
expected to search his memory or the context for the relevant referent.
Inasmuch as proper names lack descriptive content, do they have all of
the flexibility of use exhibited by descriptions? What would count as a
nonliteral use of a proper name? Perhaps using a name to refer to
something it was not the name of. Of course, unless the connection
between the object and its pseudoname can be inferred, communica-
tion will break down. But if, say, Sand H mutually believe that Sam
considers himself to be something of a renaissance man, S might utter
"Here comes Leonardo," intending H to infer that S is referring to
Sam.
We have completed our schematic look at the inference from what is
meant bye to what is said. We have made no attempt to deal with the
various technical problems generated by particular sorts of referring
expressions, such as definite descriptions, personal pronouns, demon-
strative pronouns, temporal adverbs, and proper names. Our concern
here has been simply to layout the general pattern of inference from
what is meant by e to what is said. H's identification of the operative
meaning of e fixes the operative senses, the predications, and the sen-
tence type. From these, together with mutual contextual beliefs, H is to
identify what S said. That is a matter of determining the type of saying
(from the sentence type), time specifications, 6 and what referents under
which descriptions are ascribed the properties (and relations) predi-
cated.
Locutionary Acts 29
2.4. SAYING THAT
There are three constraints on a correct account of L3 in the SAS: (i)
the account must render L3 determinable from L2; (ii) since L3 repre-
sents the hearer's identification of what the speaker has said, the ac-
count should accord with how "said that" is commonly ascribed; and
(iii) it must provide an adequate basis for the hearer to reach line L4 of
the SAS ("S, if speaking literally, is F*-ing that p"). We have seen
already how (i) is met and in the next section the SAS will be developed
so as to meet (iii). But before going further, we should look into (ii) in
order to clarify how the hearer specifies what the speaker says.
One problem is that there are two ways of taking said-that sentences:
as referentially opaque or as referentially transparent. When such a
sentence is taken opaquely, it makes a difference which referring ex-
pressions are used within the scope of "said that." For example, if S
utters "The inventor of Yo-Yos died happy," thereby saying that the
inventor of Yo-Yos died happy, it is false, on the opaque construction,
that S said that the inventor of parking meters died happy, even though
the inventor of Yo-yos also invented parking meters. Even if H were
aware of this curious fact, he would not regard S as having said that the
inventor of parking meters died happy - though that is nevertheless
true if "said-that" is taken transparently. Fortunately, ambiguity can
be avoided if "said of" is used instead of the transparent "said that."
Thus H could describe S as having said of the inventor of parking
mete;s (whom S referred to as the inventor of Yo-Yos) that he died
happy.
The SAS seems to require that" said that" be taken opaquely. Sup-
pose the speaker utters "The man with a martini is a famous poet,"
where the description is being used to pick out a specific man (S has
him in sight). If the mal) (with a martini) is in fact an unknown poet, it
would be incorrect to use (6a) , taken opaquely, rather than (6b), to
report what S said.
(6) a. S said that the obscure poet with a martini is a famous poet.
b. S said, of the obscure poet with a martini, that he is a famous
poet.
For the purposes of the SAS however, (6b) is unhelpful in that it does
not reflect S's point of view, which H is presumably trying to repre-
The Theory 30
sent. Accordingly, we should restrict the specification of what is said
by letting the sense of the referring expressions play a role in deter-
mining what is said. Thus "the present king of Sweden" could be
replaced by "the present male monarch of Sweden" but not by "King
Gustav." The restriction of H' s ascription of" said that" to the opaque
interpretation not only seeks to capture S's point of view, it records
information that could well affect illocutionary force. Compare (6a)
with (6c):
(6) c. S said that the schmuck over there with a martini is a famous
poet.
In general, H determines what is said by identifying the operative
meanings of the predicates, the operative senses of the referring ex-
pressions, the time specifications and the referents (if any), together
with sentence type. Problem cases come to mind, many of which are
presently the subject of intense philosophical inquiry. Since judgments
on them are varied (both between persons and between cases), the
following remarks should not be taken as conclusive. Indeed, we will
find some special cases in which it is not clear at all what the speaker
said.
False Descriptions
Again suppose S utters "The man with a martini is a famous poet." But
this time, unbeknownst to S, the man's glass contains nothing but
water. Surely S said of that man that he is a famous poet, but did S say
that the man with a martini is a famous poet? So far, the theory does
not predict anything specific because we have not yet defined the
operative senses of referring expressions. Are they the descriptive
content of the referring expressions? The descriptions S believes are
true of the referents? Descriptions S believes H thinks are true of the
referents? We take "said that" to report the senses of the referring
expressions (in this case the descriptive content of the description) as
well as the things referred to by S with each of these expressions, but it
is not required that these senses determine the objects referred to. Thus
S is free to pick whatever referring expressions best suit his immediate
conversational objectives; in particular, he is free to pick descriptions
he believes to be false of the referents. He might believe, for example,
that H believes them to be true of the referents. Or he might have
reason to expect that H, though also believing them to be false of the
referents (perhaps also believing S to believe that), will make the right
. Locutionary Acts 31
identifications. In such case H can still use the description to identify
the referents, and nothing in the schema precludes this. Indeed, the
schema demands this. Consider the case where S utters the sentence
about the man with a martini, believing that the man is drinking water
but thinking H believes it is a martini. If S does not want to bother
mentioning that the man is drinking water, S can say what he wants to
say by exploiting H's false beliefs. In general, the requirement is that
H, in representing what S says, pick out the referents in the way he
thinks he is intended to.
Proper Names
Suppose that speaker S utters the sentence "The Morning Star is really
a planet." Did S say that Venus is really a planet? Suppose S utters the
sentence "Venus is really a planet." Did S say that the Morning Star is
really a planet? Our account can make no prediction without an ac-
count of the sense of proper names and a way of determining which
expressions are functioning as proper names. Is H to infer that "the
Morning Star" is functioning as a proper name? As a description? As
both? And does the language, the speaker, or both determine this?
These are hard questions and we have no general doctrine of proper
names. However one thing does seem relatively clear. If by "sense" of
a referring expression one means 'descriptive content," then proper
names do not have sense. But if by "sense" one means 'that which
determines reference,' then nonvacuous names would appear to have
sense-after all, something must determine their reference; surely
there is some sort of connection between such words and things. 7
So far we have left "sense" open between these two interpretations
because it was not clear which conception is required by the ordinary
use of "said that." In order to proceed we will adopt the safest policy,
one that licenses an identification of what was said only under the most
restricted circumstances. The schema may fail to reflect some legiti-
mate inferences, but it will rarely legitimate bad ones. We will assume
that proper names have no descriptive content but that they do have
something that contributes to their reference potential, and we will call
this their sense. However, since what connects a name to an object will
usually involve the name itself (there being no meaning), reporting
what was said in the utterance of a proper name will require use of the
same name or one with the same connections to the referent for the
speaker. Any shift in the report from one name to a name not con-
nected in the same way to the same object for the original speaker will
The Theory 32
be a change in what is said. Thus, the answer to our questions regarding
"Venus" and "the Morning Star" is no.
In our earlier Leonardo example, it is unclear what the speaker said
in uttering "Here comes Leonardo." If he did not say that there came
Leonardo (he was referring to Sam), did he say that there came Sam?
Vacuous Terms
Suppose that speaker S utters "The largest prime number is larger than
100." Since nothing satisfies the description "largest prime number"
and there is no contextual clue as to what else might be referred to,
what has S said, if anything at all? The same question can be posed for
vacuous proper names. Suppose that S utters" Santa Claus lives at the
North Pole." What did S say, if anything?
Our account does not yet predict these cases. Furthermore, it is not
clear how determinate our pretheoretic use of "said-that" is in such
cases. However, if what is said can be a bearer of truth values, then it
might be supposed that what was said with the vacuous description can
be specified in terms of Russell's (1905) theory of descriptions: there is
just one number that is the largest prime and it is larger than 100. How
about vacuous proper names? Both ordinary usage and the theory of
truth leave that case unsettled.
Pronouns
Suppose that S utters "He is a conservative" referring to William F.
Buckley. Did S say that Buckley is a conservative? Suppose S utters
"William F. Buckley is a conservative." Did S say that he (referring to
Buckley) is a conservative? Surely in the first case S said of Buckley
that he is a conservative, but did S say that Buckley is a conservative?
On the assumption that pronouns like "he," "she," "them," "it,"
and so on, do have (minimal) descriptive content as their sense, our
account predicts that such shifts from proper names to pronouns
change what is said. So the answer to our questions is no.
Mistaken Linguistic Beliefs
Suppose that speaker S utters "The milkman is erotic" thinking
"erotic" means 'erratic.' Did S say that the milkman was easily aroused
(or whatever)? We think not, but this causes no problem for the
schema. Although the linguistic basis for the inference to L2 of the SAS
is intended to be the hearer's (shared) knowledge of the language, in
Locutionary Acts 33
this case the hearer cannot succeed on this linguistic basis alone. In this
case the hearer must rely on another fact as well, that S thinks" erotic"
means 'erratic.' What the speaker said is determined as usual by the
operative meaning of what was uttered, although that was not one of its
meanings in L.
Mistaken Factual Beliefs
Suppose that speaker S, thinking that whales are fish and not mammals,
utters "That's not a boat, it's a huge fish that is attacking." Even if H
knows better, he could still rightly regard S as having said that a whale
was coming.
Nonliteral Utterances
Suppose that S utters "Mac is a fine friend" with obvious insincerity.
Did S say that Mac is a fine friend? The answer is complicated. If there
is no intonational difference between this utterance and the sincere
case, then it seems that S did say that Mac was a fine friend, though
again he did not mean it, does not believe it, would not be fairly rep-
resented as having committed himself to the truth of it. If, on the other
hand, there is an intonational clue to the sarcastic reading, it seems that
such an utterance means the opposite of what it means without the
change of intonation, and so the speaker may well have said that Mac
was a scoundrel (or whatever). We see no reason to deny that there are
characteristic sarcastic intonation contours with semantic effects.
Slips of the Tongue
Suppose speaker S utters "The vote is anonymous," intending to say
that the vote is unanimous. Did S say that the vote is anonymous? Yes,
but surely, as one might ordinarily remark, he did not mean it (that is,
intend to say it). It would be misleading but true to report S as saying
that the vote is anonymous without adding some rider to the effect that
S intended to say that the vote is unanimous, but has misspoken. For
the purposes of the SAS, at any rate, the hearer could make the appro-
priate adjustment if he recognized what S intended to utter.
Illocutionary Adverbials
An interesting problem for our account of saying is illustrated by an
utterance of "Frankly, Bruckner bores me." S is saying at least that
Bruckner bores him, but we cannot readily describe him as saying that
The Theory 34
frankly Bruckner bores him. The reason is that "frankly" does not
modify the content of what S is saying but rather his assertion of it. We
discuss such illocutionary adverbials in chapter 10.
We may conclude that except for the troublesome cases just men-
tioned, which generate special problems to be taken up later, what the
hearer identifies as represented at line L3 of the SAS is just what would
ordinarily be described as "what the speaker said."
2.S. DETERMINING LITERAL ILLOCUTIONARY FORCE
POTENTIAL FROM LOCUTIONARY ACTS
At L3 of the SAS, H has inferred that S is saying that *( ...p ... ). How
does H infer L4 from L3?
L3. S is saying that *( ...p ... ).
L4. S, if speaking literally, is F*-ing that p.
In other words, what can H infer about any illocutionary act being
literally performed in the utterance of e? If S is speaking literally, then
what S says delimits what his illocutionary act can be, but in general it
does not fully determine what that act is. Indeed, it is possible (if the
communicative presumption is not in effect) for S to be performing no
illocutionary act at all.
In section 1.4 we characterized an illocutionary act as performed
literally when the (propositional) content of the locutionary act and
illocutionary act are the same, and the illocutionary act type is L-
compatible with the sentence type and meaning of e. Because of the
connection we have established in section 2.3 between the sentence
type and meaning of e and the locutionary act performed in uttering e,
we can view L-compatibility simply as a relation between illocutionary
and locutionary acts. A certain type of illocutionary act is L-compatible
with a given locutionary act only when S has the right sort of communi-
cative intent. Specifically, it is required that
Compatibility Condition (CC):
i. If S is saying that I-( ...p ... ), S is expressing the belief that p;
ii. If S is saying that !( ...p ... ), S is expressing the desire that H make it
the case that p; and
iii. If S is saying that ?( ...p ... ), S is expressing the desire that H tell S
whether or not p.
Locutionary Acts 35
We will use the phrase "s is expressing attitude A *( ... p ... ) to sche-
matize the connections set down in this definition. In chapter 3 we
propose a detailed taxonomy of communicative illocutionary acts in
terms of types of expressed attitudes. Various subtle distinctions are
drawn there, but the general types of expressed attitudes mentioned in
the compatibility condition will suffice for present purposes. We can
say that an illocutionary act type F is L-compatible with a given locu-
tionary act of saying that *( ...p ... ) if the attitude expressed in F-ing, as
specified in the taxonomy, meets the compatibility condition. From this
we get a natural definition of the literal performance of an illocutionary
act:
Literal Performance (Lit): S's F-ing that P in saying that *( ...p ... )
is literal just in case:
i. P = ( .. .p .. . ), and
ii. F-ing is L-compatible with saying that *( ...p ... ).
In the schema it is convenient to represent literally performed illocu-
tionary acts with the notation: F*-ing that p. The hearer's inference
from L3 to L4 of the SAS can accordingly be represented for the case
of declarative utterances, for example, as follows:
L3. S is saying that 1-( .. .p .. . ).
(a) If S is speaking literally, S is expressing the belief that p.
L3, CC (i)
(b) If S is speaking literally, S is asserting that p, or otherwise
F*-ing that p.
L3(a), Lit
L4. S, if speaking literally, is asserting that p.
L3(b), CP, MCBs
It should be clear that L4 might well have different content if H had a
different hypothesis at L3(b), for example, the hypothesis that S was
making a suggestion rather than an assertion: different types of illocu-
tionary acts could be literally performed in saying that *( ...p ... ).
Consider some sample sentences. Suppose that:
(7) a. In uttering "John will close the door" S is saying that it will be
the case that John closes the door.
b. In uttering "Close the door" S is saying that H is to close the
door.s
The Theory 36
Locutionary Acts 37

c. In uttering "Did John close the door?" S is asking (or saying that (8) a. I intend to repay you.
H is to tell S) whether or not John closed the door. b. It will be the case that I repay you.
Looking ahead to the taxonomy we develop in the next chapter, we can c. I intend to get even with you.
see, considering sentence type alone, which illocutionary acts are L- d. It will be the case that I get even with you.
compatible with each of these sayings. In particular, only certain con-
Our original sentence is at best constative-or-commissive determinate
statives 9 and commissives are L-compatible with S's saying (7a), only in virtue of the fact that each of its readings is compatible with a variety
certain directives are L-compatible with S's saying (7b) , and only of constatives and a variety of commissives.
questions (a subtype of directives) are L-compatible with S's saying
(7c). Considerations involving propositional content (what is said) 2.6. CONCLUSION
further narrow down the set of L-compatible illocutionary acts. Per-
forming a predictive, not a retrodictive, is L-compatible with (7a) ,
We have spelled out the pattern of inference H makes in order to
because a predictive requires future time reference, a retrodictive past identify what S says in uttering something and thereby to determine its
time reference. Some commissives (for example, acts of swearing that)
literal illocutionary force potential. Schematized as a whole, this phase
are L-compatible with (7a), but others (such as surrendering, inviting, of the inference goes as follows:
bidding, and volunteering) are not. Having inferred that S is saying that
it will be the case that John closes the door, H is still free (linguistically) Basis
to infer that S is doing anyone of a number of different illocutionary Ll. S is uttering e. hearing Sutter e
acts. These include predicting, guessing, informing, confirming, con- (a) e means ... and __ in L. Ll, knowledge of L
ceding, assenting, replying, suggesting, and guaranteeing. To infer (b) S means ... or __ bye. Ll(a), LP
which one S was performing, H needs more information than is avail- (c) The supposition that S
able from what S has said. These remarks hold true, mutatis mutandis, means __ bye is contex-
for (7b) and (7c) as well. L-compatible with S's saying thatH is to close tually less appropriate. Ll(b), MCBs
the door are requesting, demanding, and ordering, but not stating, pro- L2. S means ... bye. Ll(b), Ll(c)
hibiting, promising, or congratulating. Moreover, questioning, query- (a) e contains referring ex-
ing, and inquiring, but virtually no other illocutionary act named in our pression(s) R, knowledge of L
taxonomy, are L-compatible with S's asking H whether or not John (b) In using R, S is referring to
closed the door. Finally, it should not be thought that only word order the such-and-such(s). L2, L2(a), MCBs
and propositional content contribute to compatibility, for such things L3. S is saying that *( ...p ... ). L2, L2(b), LP
as performative verbs and intonation contribute as well. We have very (a) If S is speaking literally, S
little to say at the moment about intonation,IO and we discuss the spe- is expressing attitude A
cial problem of performatives in chapter 10. *( ...p .. . ). L3, CC
We will call the linguistic side of L-compatibility Jorce-determinacy (b) If S is speaking literally, S
(F-determinacy). An expression e is F-determinate with respect to a is F*-ing that p, or ... L3(a), Lit
particular type F of illocutlonary act just in case if S utters e and L4. S, if speaking literally, is
performs some illocutionary act literally, then S is F-ing. Notice that a F*-ing that p. L3(b), CP, MCBs
sentence, on a reading, can be more F-determinate than the sentence
The substeps of these inferences need not be specifically intended by S.
taken simpliciter. Consider the sentence "I will pay you back (for
Rather, they comprise typical patterns of inference that hearers actu-
that)." F-determinacy depends on meaning, but the operative meaning
ally make. Only the numbered lines figure in what S needs to intend if
of this sentence can be any of the following:
he is to communicate linguistically.
The Theory 38
In the first section of this chapter, we were left with a certain ambi-
guity in Austin's characterization oflocutionary acts, as exemplified by
the phrase, "uttering a sentence with a certain sense and reference." At
L2 of the schema, what S means by e is one of the meanings of e in L,
as specified at Ll(a). At this stage uttering e with a certain sense and
reference is merely to intend a certain one of its meanings to be opera-
tive and for the referents of its referring expressions to be delimited
accordingly. Nothing is yet entailed about what the speaker is saying,
which is represented by L3. Only there is it inferred what S is referring
to, generally in a way much more determinate than the way referring
expressions refer (as a consequence of their senses). So we can resolve
the ambiguity in Austin's formulation by letting locutionary acts be
represented by L3. Whatever else fits his ambiguous characterizations
falls under earlier steps or substeps of the SAS and consequently gets
accounted for without being included in the locutionary act itself.Ii
Certain philosophical and linguistic issues remain. For instance,
what are the different types of meaning alluded to in the steps from Ll
to L2, and how are they to be specified? What is the nature of such
phenomena as implication and presupposition, which are closely con-
nected to what is said but are not part of what is said? These issues will
be taken up in chapter 8. In the next chapter we develop our taxonomy
of communicative illocutionary acts.
Chapter Three A Taxonomy of
Communicative
Illocutionary Acts
Types of illocutionary acts are distinguished by types of illocutionary
intents (intended illocutionary effects). Since illocutionary intents are
fulfilled if the hearer recognizes the attitudes expressed by the speaker,
types of illocutionary intents correspond to types of expressed atti-
tudes. Accordingly, we will classify types of illocutionary acts in terms
of types of expressed attitudes. This will enable us to integrate our
taxonomy with the SAS.
To express an attitude in uttering something is, in our conception, to
R-intend that the hearer take one's utterance as reason to believe one
has the attitude. The speaker need not have the attitude expressed, and
the hearer need not form a corresponding attitude. The speaker's hav-
ing the attitude expressed is the mark of sincerity, but illocutionary or
communicative success does not require sincerity. If the hearer forms a
corresponding attitude that the speaker intended him to form, the
speaker has achieved a perlocutionary effect in addition to illocution-
ary uptake.
Individuating communicative illocutionary acts in terms of expressed
attitudes leaves ample room for a rich diversity of act types. In most
cases the speaker expresses not only his own (putative) attitude toward
the propositional content but also his intention that the hearer form a
corresponding attitude. For example, to inform someone of something
is not only to express a belief in it but also to express one's intention
that the hearer believe it. Act types are further differentiated by the
reasons for or the strengths of the attitudes expressed. For example,
what we call "confirmatives" are distinguished from assertions gener-
ally by S's expressing his belief as being the result of some truth-
seeking procedure. And within the class of what we call "advisories,"
the difference between urging someone to do something and merely
The Theory 40 Communicative Illocutionary Acts 41

suggesting he do it is marked by the difference in strength in S's ex- Communicative Illocutionary Acts
pressed intention or desire that H do it. Finally, some act types are
picked out by expressed attitudes concerning the context or occasion
of utterance: an answer is R-intended to be taken as a response to a I II III IV
question; an apology is R-intended to be taken as occasioned by the
Constatives Directives Commissives Acknowledgments
speaker's having done some regrettable thing to the hearer. As we
develop the taxonomy in detail, just how these various dimensions of Assertives Requestives Promises Apologize
Predictives Questions Offers Condole
expressed attitudes determine illocutionary act types will be made clear Retrodictives Requirements Congratulate
and concrete. Descriptives Prohibitives Greet
Many taxonomies of illocutionary acts have been proposed, but we Ascriptives Permissives Thank
Informatives Advisories Bid
will not discuss or compare all of them. Austin's original scheme (1962, Confirmatives Accept
Lecture XII) included a rich variety of illocutionary act types, but, as Concessives Reject
Searle (1975b) has argued, there are no clear principles by which Austin Retractives
Assentives
collected them into his five classes. All subsequent taxonomies! are Dissentives
attempted improvements on Austin's, but only Searle's is tied to a Disputatives
general theory of illocutionary acts. We agree with Searle that a scheme Responsives
Suggestives
of classification should be principled. Its categories should not over- Suppositives
lap2-at least not beyond what can be expected from the nature of the
subject-and the entries in each category should satisfy the criteria for Figure 3.1 Classification of communicative illocutionary acts (in category IV
belonging to that category. Moreover, to be of theoretical interest the specific verbs are listed)
scheme's bases of classification must be tied to some systematic ac-
count of illocutionary acts. positives, exercitives, commissives, and behabitives, respectively, and
The fundamental idea behind our taxonomy is that the illocutionary closely to Searle's representatives, directives, commissives, and ex-
intents, or expressed attitudes, by which types ofillocutionary acts are pressives, although our characterizations of them are different from
distinguished are all homogeneous with the speech act schema. That is, Searle's.
the SAS represents the general form of illocutionary intention and in- For us, constatives express the speaker's belief and his intention or
ference, and the entries in the taxonomy provide the content, as is desire that the hearer have or form a like belief. Directives express the
evident in the concluding step of the SAS: the identification of the speaker's attitude toward some prospective action by the hearer and
illocutionary act being performed. Since such acts are identified by his intention that his utterance, or the attitude it expresses, be taken as
their intents (H's recognition of S's expressed ~ttitudes), the distin- a reason for the hearer's action. Commissives express the speaker's
guishing features of each illocutionary act type specify the very thing H intention and belief that his utterance obligates him to do something
must identify in the last step of the SAS. (perhaps under certain conditions). And acknowledgments express feel-
A more obvious merit (we hope) of our taxonomy is its comprehen- ings regarding the hearer or, in cases where the utterance is clearly
siveness and explicitness. It covers a great many types of illocutionary perfunctory or formal, the speaker's intention that his utterance satisfy
acts in detail, not only labeling them but specifying what distinguishes a social expectation to express certain feelings and his belief that it
them. We divide illocutionary acts into six general categories. Two of does. Figure 3.1 lists the subcategories falling under these four head-
these categories, effectives and verdictives, are conventional not,com- ings. They will be discussed in detail in the sections to follow, where
municative; they will be discussed in chapter 6. The four main kinds of specific R-intentions will be spelled out, together with, when not obvi-
communicative illocutionary acts are constatives, directives, commis- ous, the correlative perlocutionary intentions.
sives, and acknowledgments; these correspond roughly to Austin's ex-
The Theory 42
3.1. CONSTATIVES
In general, a constative is the expression of a belief, together with the
expression of an intention that the hearer form (or continue to hold) a
like belief. The following analyses of various specific kinds of con-
statives exhibit this pattern.
Assertives (simple): (affirm, allege, assert, aver, avow, claim, declare,
deny (assert ... not), indicate, maintain, propound, say, state, submit)
In uttering e, S asserts that P if S expresses:
i. the belief that P, and
ii. the intention that H believe that P.
Predictives: (forecast, predict, prophesy)
In uttering e, S predicts that P if S expresses:
i. the belief that it will be the case that P, and
ii. the intention that H believe that it will be the case that P.
Retrodictives: (recount, report)
In uttering e, S retrodicts that P if S expresses:
i. the belief that it was the case that P, and
ii. the intention that H believe that it was the case that P.
Descriptives: (appraise, assess, call, categorize, characterize, classify,
date, describe, diagnose, evaluate, grade, identify, portray, rank)
In uttering e, S describes 0 as F if S expresses:
i. the belief that 0 is F, and
ii. the intention that H believe that 0 is F.
Ascriptives: (ascribe, attribute, predicate)
In uttering e, S ascribes F to 0 if S expresses:
i. the belief that F applies to 0, and
ii. the intention that H believe that F applies to o.
Informatives: (advise, announce, apprise, disclose, inform, insist, no-
tify, point out, report, reveal, tell, testify)
In uttering e, S informs H that P if S expresses:
i. the belief that P, and
ii. the intention that H form the belief that P.
Confirmatives: (appraise, assess, bear witness, certify, conclude, con-
firm, corroborate, diagnose, find, judge, substantiate, testify, validate,
verify, vouch for)
Communicative Illocutionary Acts 43
In uttering e, S confirms (the claim) that P if S expresses:
i. the belief that P, based on some truth-seeking procedure, and
ii. the intention that H believe that P because S has support for P.
Concessives: (acknowledge, admit, agree, allow, assent, concede, con-
cur, confess, grant, own)
In uttering e, S concedes that P if S expresses:
i. the belief that P, contrary to what he would like to believe or con-
trary to what he previously believed or avowed, and
ii. the intention that H believe that P.
Retractives: (abjure, correct, deny, disavow, disclaim, disown, recant,
renounce, repudiate, retract, take back, withdraw)
In uttering e, S retracts the claim that P if S expresses:
i. that he no longer believes that P, contrary to what he previously
indicated he believed, and
ii. the intention that H not believe that P.
Assentives: (accept, agree, assent, concur)
In uttering e, S assents to the claim that P if S expresses:
i. the belief that P, as claimed by H (or as otherwise under discussion),
and
ii. the intention (perhaps already fulfilled) that H believe that P.
Dissentives: (differ, disagree, dissent, reject)
In uttering e, S dissents from the claim that P if S expresses:
i. the disbelief that P, contrary to what was claimed by H (or was
otherwise under discussion), and
ii. the intention that H disbelieve that P.
Disputatives: (demur, dispute, object, protest, question)
In uttering e, S disputes the claim that P if S expresses:
i. the belief that there is reason not to believe that P, contrary to what
was claimed by H (or was otherwise under discussion), and
ii. the intention that H believe that there is reason not to believe that P.
Responsives: (answer, reply, respond, retort)
In uttering e, S responds that P if S expresses:
i. the belief that P, which H has inquired about, and
ii. the intention that H believe that P.
Suggestives: (conjecture, guess, hypothesize, speculate, suggest)
In uttering e, S suggests that P if S expresses:
The Theory 44
i. the belief that there is reason, but not sufficient reason, to believe
that P, and
ii. the intention that H believe that there is reason, but not sufficient
reason, to believe that P.
Suppositives: (assume, hypothesize, postulate, stipulate, suppose, the-
orize)
In uttering e, S supposes that P if S expresses:
i. the belief that it is worth considering the consequences of P, and
ii. the intention that H believe that it is worth considering the con-
sequences of P.
We said at the outset that in general, constatives are the expression
of a belief, together with the expression of an intention that the hearer
form, or continue to hold, a similar belief. Simple assertives, descrip-
tives, and ascriptives are of this sort. The perlocutionary intention nor-
mally accompanying these acts is that the hearer believe, or continue to
believe, the proposition (P) in question, perhaps by way of believing
that the speaker believes it. That is, over and above identifying the
belief and the intention expressed, H is intended to believe that S
believes the proposition and, possibly because of this, to believe the
proposition himself. Of course, there are cases in which the speaker
knows perfectly well that the hearer disbelieves that P and will not
change his mind just because S believes that P. Nevertheless, S wants
H to ascribe the belief that P to S, and, even if he does not intend H to
believe that P, at least he wishes that H believe it; in these cases S
expresses the wish, rather than the intention, that H believe that P. As
we suggested in chapter 1, when in performing an illocutionary act S
expresses a certain intention regarding H, in general he has the cor-
responding perlocutionary intention. However, if S disbelieves that his
utterance will have any such perlocutionary effect on H, he expresses
at most the wish that such an effect result. So S may have the perlocu-
tionary intention that H attribute to him this wish. Finally, there are
cases where S thinks H won't take his utterance as sincere. That is, S
expects H not to attribute to him the belief and the intention S is
expressing. In this case S cannot expect to have fulfilled, and therefore
cannot reasonably form, the perlocutionary intention that H believe
that he (S) believes that P, much less the further intention that H
believe it himself.
The assertives listed vary in strength of belief expressed and in
the corresponding expressed intention. When one maintains or avows
Communicative Illocutionary Acts 45
something, one's expressed belief and intention are very strong,
whereas the belief and intention expressed when one alleges or submits
that something is the case are much weaker. We have reserved the
separate categories of suggestives and suppositives for constative utter-
ances that express not even a weak belief that P, but only the belief that
there is reason to believe that P or that (because it is possible or
plausible that P is true) it is worth considering the consequences of P.
Some of the verbs listed as descriptives are rather specialized in their
coverage. Verbs like "appraise," "date," "diagnose," and "grade"
apply to restricted categories of things. These verbs are not synony-
mous, but that does not mean they designate different sorts of illocu-
tionary acts. They don't: not every difference between illocutionary
verbs is illocutionary.
What distinguishes informatives from simple assertives is that the
speaker expresses (in addition to his belief) the intention that the hearer
form the belief that P. For assertives, S's expressed intention is that H
form the belief, or continue to believe, that P. We might say that at the
time of utterance, S presumes that H does not believe that P. Here we
rely on a notion of illocutionary presumption: in his illocutionary inten-
tion, S presumes that q if the truth of q is necessary for the rationality
of his illocutionary intention. Of course, the notion of rationality is
multifarious; all we mean, in speaking of the rationality ofillocutionary
intentions, is that there is good reason to believe that the intention will
be fulfilled (by being recognized). Generally, for an illocutionary act
with a presumption, the truth of that presumption is necessary for the
success of that act. With this rough notion of illocutionary presump-
tion, we will be able to distinguish some of the other kinds of con-
statives partly in terms of what is presumed.
Concessives, retractives, assentives, dissentives, and disputatives all
involve a presumption about the contextual relevance of the expressed
belief. A concessive expresses a belief contrary to what S would like to
believe or contrary to what he previously believed or avowed, whereas
a retractive expresses that S no longer believes what he previously
indicated he believed, but in both cases it is presumed that the question
of S' s belief has come up in the conversation or is otherwise directly
relevant to the current stage of conversation. Assentives, dissentives,
and disputatives all presume that a certain claim has been made by H or
that someone's claim, not necessarily S's or H's, is under discussion.
To assent that P is to express agreement with this claim, to dissent from
it is to express disagreement, and to dispute it is to express the belief
that there is reason not to believe that P.
The Theory 46
Confirmatives express not only the speaker's belief that P but that he
believes it as a result of some truth-seeking procedure, such as ob-
servation, investigation, or argument. Since the latter belief is also
expressed, it is not a mere illocutionary presumption. Responsives are
R-intended as replies to an inquiry by H. In responding that P, S
expresses his belief that P and that he is so doing in answer to H.
Obviously, in some contexts a responsive can be a dissentive or a
disputative as well. Suggestives and suppositives that P are not expres-
sions of belief that P. In suggesting (conjecturing, hypothesizing) that
P, S expresses merely the belief that there is reason to believe that P,
but not sufficient reason to believe it. And in supposing (assuming,
postulating) that P, what S expresses is the belief that it is worth con-
sidering the consequences of P, irrespective of whether it is true that P.
Here S is likely to have the perlocutionary intention that H is to expect
S to take up a discussion of P or its consequences.
An analysis of constative verbs with specialized ranges of appli-
cation, like "appraise," "testify," "recant," and "postulate," would
specify what that range of application is and that the speaker presumes
his utterance to fall within this range. Such an analysis would take into
account the strength of the expressed attitude. Just as, among asser-
tives, maintaining something expresses a stronger belief than alleging
it, so among disputatives, to object to something expresses a stronger
belief (regarding reasons for disbelieving the proposition in question)
than does questioning something. There is a similar difference between
the suggestives hypothesizing and conjecturing. The analyses for the
central cases of each type would have to be modified slightly to reflect
these differences.
For certain purposes, the subtypes we have given could be sup-
plemented or subdivided further. No doubt additions could be made to
our list of verbs for each type, though we suspect that most such verbs
would be too specialized in scope to be of interest here. Finally, we
should point out that some verbs occur under more than one heading.
This does not necessarily mean that the types overlap, only that some
verbs name more than one type. Nevertheless, there is such overlap.
Most of the specialized types of constatives satisfy the definition of
assertives, and responsives, for example, overlap with disputatives and
with suggestives. This means not that our definitions or conceptions of
these types of constatives are hazy, but that some illocutionary act
tokens can be of more than one type, performed with the R-intention
appropriate to each.
communicative Illocutionary Acts' 47
3.2. DIRECTIVES
Directives express the speaker's attitude toward some prospective ac-
tion by the hearer. If this were all they expressed, they would be
merely constatives with a restriction on propositional content (namely,
that a prospective action be ascribed to the hearer). However, they also
express the speaker's intention (desire, wish) that his utterance or the
attitude it expresses be taken as (a) reason for the hearer to act. Rather
than use Austin's term "exercitive," which seems somewhat restricted
in scope, we have borrowed Searle's term "directive." It is both to the
point and conveniently vague, being broad enough to cover the six
kinds of acts that belong in this category.
Requestives: (ask, beg, beseech, implore, insist, invite, petition, plead,
pray, request, solicit, summon, supplicate, tell, urge)
In uttering e, S requests H to A if S expresses:
i. the desire that H do A, and
ii. the intention that H do A because (at least partly) of S's desire.
Questions: (ask, inquire, interrogate, query, question, quiz)
In uttering e, S questions H as to whether or not P if S expresses:
i. the desire that H tell S whether or not P, and
ii. the intention that H tell S whether or not P because of S's desire.
Requirements: (bid, charge, command, demand, dictate, direct, enjoin,
instruct, order, prescribe, require)
In uttering e, S requires H to A if S expresses:
i. the belief that his utterance, in virtue of his authority over H, con-
stitutes sufficient reason for H to A, and
ii. the intention that H do A because of S's utterance.
Prohibitives: (enjoin, forbid, prohibit, proscribe, restrict)
In uttering e, S prohibits H from A-ing if S expresses:
i. the belief that his utterance, in virtue of his authority over H, con-
stitutes sufficient reason for H not to A, and
ii. the intention that because of S's utterance H not do A.
Permissives: (agree to, allow, authorize, bless, consent to, dismiss,
excuse, exempt, forgive, grant, license, pardon, release, sanction)
In uttering e, S permits H to A if S expresses:
i. the belief that his utterance, in virtue of his authority over H, entitles
H to A, and
ii. the intention that H believe that S's utterance entitles him to A.
The Theory 48
Advisories: (admonish, advise, caution, counsel, propose, recommend,
suggest, urge, warn)
In uttering e, S advises H to A if S expresses:
i. the belief that there is (sufficient) reason for H to A, and
ii. the intention thatH take S's belief as (sufficient) reason for him toA.
Requestives express the speaker's desire that the hearer do some-
thing. Moreover, they express the speaker's intention (or, if it is clear
that he doesn't expect compliance, his desire or wish) that the hearer
take this expressed desire as reason (or part of his reason) to act. The
corresponding perlocutionary intentions, as might be foreseen, are that
H take S actually to have the desire and the intention he is expressing
and that H perform the action requested of him. Verbs of requesting
connote variation in strength of attitude expressed, as between "in-
vite" and "insist" and between "ask" and "beg." The stronger ones
convey a sense of earnestness or urgency. "Beseech" and "suppli-
cate," among others, convey both an appeal to the hearer's sympathy
and a special manner of performance. Some verbs of requesting are
rather specialized in scope. "Summon" (or "invite" taken narrowly)
refer to requests for the hearer's presence; "beg" and "solicit" apply to
requests for contributions or favors.
Questions are special cases of requests, special in that what is re-
quested is that the hearer provide the speaker with certain information.
There are differences between questions, but not all of them are im-
portant for an illocutionary taxonomy. There are exam questions and
rhetorical questions. "Interrogate" suggests duress in a way that "ask"
does not. Finally, "quiz" and "query" do not quite fit our analysis, in
that they cannot be used to report the content of a question but only its
topic (S quizzed H about topology).
Requirements, such as ordering or dictating, should not be confused
with requests, even strong ones. There is an important difference. In
requesting, the speaker expresses his intention that the hearer take his
(S's) expressed desire as a reason to act; in requirements S's expressed
intention is that H take S's utterance as a reason to act, indeed as
sufficient reason to act. As a matter of fact, requirements do not neces-
sarily involve the speaker's expressing any desire at all that the hearer
act in a certain way. It might be quite clear that S couldn't care less.
Instead, what S expresses is his belief that his utterance constitutes
sufficient reason for H to perform the action. In expressing this belief
and the corresponding intention, S is presuming that he has the author-
Communicative Illocutionary Acts 49
ity over H (physical, psychological, or institutional) that gives such
weight to his very utterances.
Prohibitives, such as forbidding or proscribing, are essentially re-
quirements that the hearer not do a certain thing. To prohibit someone
from smoking is to require him not to smoke. We list prohibitives
separately because they take a distinct grammatical form and because
there are a number of such verbs. We will let the entry for prohibitives
speak for itself.
Permissives, like requirements and prohibitives, presume the speak-
er's authority. They express S's belief, and his intention that H believe,
that S' s utterance constitutes sufficient reason for H to feel free to do a
certain action. The obvious reasons for issuing a permissive are either
to grant a request for permission or to remove some antecedent restric-
tion against the action in question. It would seem, therefore, that the
speaker presumes either that such a request has been made or that such
a restriction exists. It is not necessary but it is common, at least with
noninstitutional permissives, that the speaker express that he does not
wish, desire, or expect the hearer not to perform the action in question.
But, as with requirements, it is not the speaker's expressed attitude but
his utterance that is intended to figure in the hearer's reason. Some of
the verbs of permitting are highly specialized, such as "bless," "dis-
miss" ('permit to leave'), "excuse" ('permit not to make restitution'),
and "release" ('permit not to fulfill an obligation').
As for advisories, what the speaker expresses is not the desire that H
do a certain action but the belief that doing it is a good idea, that it is in
H's interest. S expresses also the intention that H take this belief of S' s
as a reason to act. 3 The corresponding perlocutionary intentions are
that H take S to believe that S actually has the attitudes he is express-
ing and that H perform the action he is being advised to perform. (It is
possible, of course, that S really does not care.) Advisories vary in
strength of expressed belief. Compare suggesting with admonishing.
Furthermore, some advisories imply a special reason that the recom-
mended action is a good idea. In warning, for example, S presumes the
presence of some likely source of danger or trouble for H.
3.3. COMMISSlVES
This is the one category of illocutionary acts for which Austin's original
label has been retained universally. Commissives are acts of obligating
oneself or of proposing to obligate oneself to do something specified in
The Theory 50
the propositional content, which may also specify conditions under
which the deed is to be done or does not have to be done. In committing
oneself to do A, one expresses the intention to do A and the belief that
one's utterance commits one to doing it, at least under the conditions
specified or mutually believed to be relevant. These conditions may
include H's accepting one's proposal or commitment to do A or at least
his not rejecting it (ordinarily, the absence of explicit rejection may be
taken as-is mutually believed to count as-acceptance). In addition
to expressing such intention and belief, the speaker expresses the in-
tention that H take him to have this intention and belief. The corre-
sponding perlocutionary intention is that H believe S has this intention
and belief and that H himself believe that S is obligated to do A, at least
if the required conditions are met.
We distinguish two main types of commissives, promises and offers.
Promises are acts of obligating oneself; offers are proposals to obligate
oneself. Under promising, we provide a sampling of special cases, in-
cluding contracting and betting, three commissive/constative hybrids
(swearing, guaranteeing, and surrendering), and one commissive/direc-
tive hybrid, inviting. The definitions are self-explanatory. As for offers,
besides the general case we give but two special cases, volunteering
and bidding. 4
Promises: (promise, swear, vow)
In uttering e, S promises H to A if S expresses:
i. the belief that his utterance obligates him to A,
ii. the intention to A, and
iii. the intention that H believe that S's utterance obligates S to A and
that S intends to A.
contract: Sand H make mutually conditional promises; fulfillment of
each is conditional on the fulfillment of the other.
bet: S promises to do something (for instance, pay a certain amount) if a
certain event occurs, on condition that H promises to do a certain thing
if a certain other event occurs.
swear that: S asserts (constative) that P and promises that he is telling
the truth.
guarantee that: S affirms (constative) the quality of something, x, and
promises to make repairs or restitution if x is relevantly defective.
guarantee x: S promises to make repairs or restitution if x is defective in
some relevant respect.
Communicative Illocutionary Acts 51
surrender: S admits (constative) defeat and promises not to continue
fighting.
invite: S requests (directive) H's presence and promises acceptance of
his presence.
Offers: (offer , propose)
In uttering e, S offers A to H if S expresses:
i. the belief that S's utterance obligates him to A on condition that H
indicates he wants S to A,
ii. the intention to A on condition that H indicates he wants S to A, and
iii. the intention that H believe that S's utterance obligates S to A and
that S intends to A, on condition that H indicates he wants S to A.
volunteer: S offers his services.
bid: S offers to give something (in a certain amount) in exchange for
something.
3.4. ACKNOWLEDGMENTS
Acknowledgments, as we call them, are the central cases of Austin's
motley class of "behabitives." They express, perfunctorily if not gen-
uinely, certain feelings toward the hearer. These feelings and their
expression are appropriate to particular sorts of occasions. For exam-
pie, greeting expresses pleasure at meeting or seeing someone, thank-
ing expresses gratitude for having received something, apologizing
expresses regret for having harmed or bothered the hearer, condoling
expresses sympathy for H's having suffered some misfortune (not S's
doing), and congratulating expresses gladness for H's having done or
received something noteworthy. Commonly, but not necessarily, such
an occasion, when it arises, is mutually recognized by Sand H, and
then it is not only appropriate but expected by H that S will issue the
relevant acknowledgment.
Because acknowledgments are expected on particular occasions,
they are often issued not so much to express a genuine feeling as to
satisfy the social expectation that such a feeling be expressed. In our
list of acknowledgments the disjunctive definitions reflect this fact.
Apologize:
In uttering e, S apologizes to H for D if S expresses:
i. regret for having done D to H, and
ii. the intention that H believe that S regrets having done D to H, or
The Theory 52
i. the intention that his utterance satisfy the social expectation that one
express regret for having done something regrettable like D, and
ii. the intention that H take S's utterance as satisfying this expectation.
Condole: (commiserate, condole)
In uttering e, S condoles H for (misfortune) D if S expresses:
i. sympathy with H's having (or suffering) D, and
ii. the intention that H believe that S sympathizes with H's having D,
or
i. the intention that his utterance satisfy the social expectation that one
express sympathy for misfortunes like D, and
ii. the intention that H take S's utterance as satisfying this expectation.
Congratulate: (compliment, congratulate, felicitate)
In uttering e, S congratulates H for D if S expresses:
i. gladness for H's having D(-ed), and
ii. the intention that H believe that S is glad that H has D(-ed), or
i. the intention that his utterance satisfy the social expectation that one
express gladness for good fortunes like D(-ing), and
ii. the intention that H take S's utterance as satisfying this expectation.
Greet:
In uttering e, S greets H if S expresses:
i. pleasure at seeing (or meeting) H, and
ii. the intention that H believe that S is pleased to see (or meet) H, or
i. the intention that his utterance satisfy the social expectation that one
express pleasure at seeing (or meeting) someone, and
ii. the intention that H take S's utterance as satisfying this expectation.
Thank:
In uttering e, S thanks H for D if S expresses:
i. gratitude to H for D, and
ii. the intention that H believe that S is grateful to H for D, or
i. the intention that his utterance satisfy the social expectation that one
express gratitude at being benefited, and
ii. the intention that H take S' s utterance as satisfying this expectation.
"No thanks": S thanks H for offering D and rejects the offer.
Bid: (bid, wish)
In uttering e, S bids H good (happy) D if S expresses:
i. the hope that H's D will be good (happy), and
Communicative Illocutionary Acts 53
ii. the intention that H believe that S hopes that H's D will be good
(happy), or
i. the intention that his utterance satisfy the social expectation that one
express good hopes when the question of another's prospects arises,
and
ii. the intention that H take S' s utterance as satisfying this expectation.
Accept-acknowledge an acknowledgment:
In uttering e, S accepts H's acknowledgment if S expresses:
i. appreciation for H's .acknowledgment, and
ii. the intention that H believe that S appreciates H's acknowledgment,
or
i. the intention that his utterance satisfy the social expectation that one
express appreciation of an acknowledgment, and
ii. the intention that H take S's utterance as satisfying this expectation.
"You're welcome": S accepts H's thanks.
Reject: (refuse, reject, spurn)
In uttering e, S rejects H's acknowledgment if S expresses:
i. lack of appreciation of H's acknowledgment,
ii. the intention that H believe that S fails to appreciate H's acknowl-
edgment, and (perhaps also)
iii. the intention that his utterance violate the social expectation that
one express appreciation of an acknowledgment, and
iv. the intention that H take S' s utterance as violating this expectation.
When one apologizes to someone, either one expresses regret (for
what one has done) or one expresses the intention that one's utterance
satisfy the social expectation to express regret (without actually ex-
pressing regret). Perfunctory acknowledgments thus require the im-
plicit cooperation of the hearer-they are issued, quite obviously to all
concerned, routinely or as a formality, as when one apologizes for
accidentally bumping someone.
Despite the fact that perfunctory acknowledgments do not express
genuine feelings, in our society they are generally regarded as acts of
courtesy. Indeed, when the acknowledgment is occasioned by some-
thing trivial or when the occasion warrants nothing more than a per-
functory acknowledgment, for the hearer to question the speaker's
sincerity would be an act of gross discourtesy and social disruptive-
ness. On the other hand, there are occasions, owing to the seriousness
 The Theory 54
of the matter or to the relation between the speaker and the hearer,
when it is expected that genuine feelings be expressed. We won't pur-
sue the sociology of acknowledgments.
In issuing an acknowledgment, the speaker presumes the existence
of the occasion to which the acknowledgment is appropriate. For
example, in thanking H for something, S presumes that he has received
something from H, and in apologizing to H, S presumes that he has
done something regrettable to H. His illocutionary act of acknowl-
edging could not succeed-the hearer could not recognize his R-
intention- unless this presumption were correct, or at least mutually
believed. The existence of the. relevant occasion is presumed, not as-
serted, by the speaker, and it is often unnecessary for him to mention
the occasion explicitly: if someone gives you a cigarette, it is enough to
say "Thank you." But if someone sends you a box of cigars, it is
necessary to say, when you next see the donor, "Thanks for the fine
cigars," or something to that effect. Condolences and congratula-
tions generally require such a specification, because they are usually
occasioned by some event removed from the current encounter of S
andH.
In acknowledgments, the only hearer-directed intention expressed
over and above the expressed feeling is that H believe that S has the
expressed feeling. Hence the only perlocutionary intention associated
with acknowledgments is that the hearer take the speaker to have the
expressed feeling or, in perfunctory cases, to regard the utterance as
satisfying the relevant social expectation. However, an acknowledg-
ment may invite an acknowledgment in response, which might be con-
strued as a perlocutionary effect if intended (it need not be, of course).
Greetings and farewells are exchanged, thanks are accepted ("You're
welcome"), congratulations and condolences are accepted with a
"Thank you" or the like, and apologies may be accepted ("That's
OK") or rejected ("Saying you're sorry isn't enough").
Similar to congratulations and condolences are biddings or (express-
ing) wishes, which may be negative, as in the case of curses. Strictly
speaking, these may be only constatives (namely, to the effect that one
has a certain wish), but in some cases biddings are called for and must
then be classed as acknowledgments.
Pardoning, excusing, and forgiving may seem to be acknowledg-
ments (asking to be pardoned, excused, or forgiven is clearly a re-
quest). However, though they may be related to acknowledgments, as
when one forgives someone for something for which he apologized (or
even excuses him from having to apologize), they seem to us to be
55
Communicative Illocutionary Acts
permissives. They are acts of releasing a person from any obligation (or
of refusing to acknowledge his putative obligation) incurred from doing
something to the speaker. Thus, they permit him not to compensate the
speaker for what he has done, or, where that is not at issue, they permit
him not to feel responsible for what he has done.
3.S. FELICITY CONDITIONS
One taxonomic issue concerns the notion of felicity conditions intro-
duced by Austin in his William James Lectures (1962). Many philos-
ophers and linguists have adopted Austin's term, but their use of it has
sometimes been rather less discriminating than his. The main problem
has been failure to observe the distinction between conditions neces-
sary (and sufficient) for the successful performance of an act, and the
conditions necessary (and sufficient) for a completely nondefective or
felicitous performance of the act. In most discussions of felicity condi-
tions, those conditions necessary for the existence of an instance of the
act are some unspecified subset of the conditions necessary for the
nondefective performance of the act. For instance, Searle (1969) gives
necessary and sufficient conditions for the (literal and direct) non-
defective performance of various speech acts, yet the absence of only
some of these conditions precludes the performance of the act. For the
sake of clarity we will call conditions that are singly necessary and
jointly sufficient for the performance of an act its success conditions;
we will call those conditions that are not success conditions but are
required for nondefectiveness felicity conditions. Is there any role for
felicity conditions to play in a theory of speech acts? If there is, how
would that role be filled in our theory?
When one looks at the literature on speech acts, there seem to be
four different motivations for having felicity conditions in a speech act
theory. First, Austin looked to the ways various acts can go wrong as a
guide to what it takes for the act to go right (1962, Lecture II). Although
he constructed a fairly elaborate taxonomy of "infelicities," his re-
peated reference to "conventional procedures" makes it pretty clear
that his doctrine of infelicities is appropriate mainly for the "highly
developed explicit performatives" associated with conventional, ritual,
and ceremonial acts, which we discuss in chapter 6. Austin's doctrine
has no obvious extension to communicative illocutionary acts.
A second motive for having felicity conditions comes from Searle
(1965, 1969), who apparently includes felicity conditions among his
necessary conditions because he is inclined to think that "we shall not
The Theory 56
be able to get a set of knock-down necessary and sufficient conditions
that will exactly mirror the ordinary use of the word 'promise'. I am
confining my discussion, therefore, to the center of the concept of
promising and ignoring the fringe, borderline, and partially defective
cases" (1965, 47). Still, it is not clear why adopting this strategy of first
analyzing paradigm cases (then seeing the rest as deviations from the
paradigm) should have the consequence that felicity conditions cannot
be distinguished from success conditions. One could just pick central
cases to give success conditions for. Of course this would involve some
modifications of Searle's format for analyzing speech acts. Essential
conditions and propositional content conditions always seem to be suc-
cess conditions, and sincerity conditions always seem to be felicity
conditions. Preparatory conditions (or parts thereof) go different ways
for different acts. Thus, a preparatory condition for asserting (that it is
not obvious to both Sand H that H knows that p) is clearly a felicity
condition, whereas part of the preparatory conditions on promising
(that S believes H would prefer S's doing A to S's not doing A) is
arguably necessary as a success condition to differentiate promising
from threatening.
A third motive for felicity conditions has come to the fore recently in
the discussion of indirect speech acts. There seem to be some generali-
zations over indirect speech acts that are best stated in terms of both
success and felicity conditions of speech acts. We will return to this
matter in chapters 4, 9, and 10.
Finally, some authors have claimed (or suggested) that felicity
conditions might be related to various grammatical phenomena. For
instance, Heringer proposes that a variety of grammatical facts con-
cerning "qualifying if-clauses," as he calls them, "can be explicated
only by reference to the illocutionary acts performed by the utterances
which contain them ... the syntactic form of the if-clause is directly
related to the intrinsic condition which it calls into question" (1972, 1).
As it turns out, though, only some of the conditions are used in this
way; in particular, they must be "conditions on the beliefs of the
speaker performing the illocutionary act" (1972, 43). Any theory taking
these beliefs into account can handle these facts if Heringer's can, and
the discussion of qualifying if-clauses does not motivate a general the-
ory of felicity conditions, at least not of the sort envisaged by Austin.
The ways an act may be defective, in an unqualified use of "defective,"
may be limited only by one's imagination. Thus it is reasonable that
only certain kinds of defect be singled out. But then some reason must
communicative Illocutionary Acts 57
be provided for why those particular kinds are theoretically significant.
So far we have found no compelling reason for a gen~ral theory. of
felicity conditions and (assuming indirect speech acts wIll not prOVIde
such a reason) propose no such theory here.
3.6. INFELICITY AND OBVIOUS INSINCERITY
Our taxonomy distinguishes types of communicative illocutionary acts
by the attitudes the speaker expresses in perfo:~ing them: the speaker
expresses a certain attitude toward the proposItional conte~t as well ~s
the intention that the hearer have or form a correspondmg proposI-
tional attitude. We have defined expressing an attitude as R-intending
the hearer to take one's utterance as reason to believe that one has the
attitude. The speaker's having the attitudes expressed is the mark of
sincerity, but sincerity is not required for communicative success; nor
is the hearer's believing the speaker has the attitudes expressed. Thus,
a communicative illocutiona~ act can succeed even if the speaker .is
insincere and even if the hearer believes he is insincere. After all, In
expressing certain attitudes the speaker is merely R-intending the
hearer to take his utterance as reason to believe him to have those
attitudes. So the speaker need not intend this reason to be sufficient,
and the hearer need not take it to be sufficient. Generally, though, it is
intended to be sufficient and is taken to be; generally it is sufficient.
Even if the speaker does not have the attitudes he is expressing, there is
no reason, most of the time, to think he does not have them. And even
if there is reason to think he does not have them, there is likely to be no
reason to think he does not R-intend one to think there is reason to
believe he has them. We may not trust him, but he may not realize that.
But suppose that the speaker's insincerity is obviously obvious; that
is Sand H mutually believe that S does not have one or another of the
a;titudes he is expressing. In the case of a statement, for example, it
might be mutually believed that S does npt believe what he is stating or
that H inalterably disbelieyes what S is ~tating. In the case of a request,
it might be mutually belieVled that S really does not want H to perform
the requested action or that H won't perform it no matter what S
wants. Such cases as these raise certain questions for our taxonomy:
(1) Does the speaker really express the attitudes it is mutually believed
he does not have? (2) Is he really performing an illocutionary act of the
sort (stating, requesting) that he would be performing if he weren't
obviously insincere?
The Theory 58
To take an example, suppose that S says to H that he (S) has not
been drinking. However, Sand H mutually believe that S has alcohol
on his breath and that an empty Ripple bottle is lying at S's feet.
Suppose that under these circumstances H cannot attribute to S the
belief that he (S) has not been drinking or the intention that H believe
that S has not been drinking.5 It is obvious to H not only that S is lying
but that S believes H believes S is lying. Assuming that H is right, does
it follow (1) that S cannot be expressing the belief, and the intention
that H believe, that S has not been drinking, and (2) that S cannot be
stating that he has not been drinking? Notwithstanding the facts of the
case, S's utterance is R-intended by S to be taken by H as reason to
think that S believes, and intends H to believe, that he has not been
drinking. Under the circumstances S cannot rationally R-intend his
utterance to be sufficient reason for H to make these attributions, but
that is not what our conception of expressing an attitude requires. By
itself, S's utterance is, and can be R-intended to be taken to be, a
reason, despite the fact that it can be overridden by mutual contextual
beliefs to the contrary. Even when defeated, a reason is a reason.
Accordingly, S can express a belief and an intention despite mutual
beliefs to the contrary. By definition, then, he can state that he has not
been drinking.
The case of obviously obvious insincerity does not present problems
for our conception of expressing an attitude or for our taxonomy of
illocutionary acts in terms of types of attitudes expressed. Indeed, the
example couldn't have been described in the way it was unless it was a
case of expressing a certain belief and a certain intention and of per-
forming the illocutionary act of stating. Otherwise, how could it be
described as a case of obviously obvious insincerity? After all, there
had to be something for the speaker to be insincere about, namely, the
attitudes he expressed. Equally, there had to be some illocutionary act
that he was performing insincerely. 6
In considering the case of obvious insincerity and its implications for
our taxonomy of communicative illocutionary acts, one should keep in
mind that all these acts are performed pursuant to the communicative
presumption (CP) and that their identity is worked out by the hearer
in accordance with the SAS. That is, the hearer must explain the
speaker's utterance by identifying the intention with which it was is-
sued, and this consists in identifying the expressed attitudes. Express-
ing an attitude is R-intending the hearer to take one's utterance as
reason, not necessarily sufficient reason, to think that one has the
Communicative Illocutionary Acts 59
attitude in question; therefore, in identifying the speaker's illocutionary
intention, the hearer must consider whether the speaker is likely to
have such an attitude. Before thinking S to be expressing attitudes that
there is mutually believed reason to believe he does not have, H might
rule out the possibility that the CP is inoperative, for example, that S is
kidding. 7 In general, any reason to think that S does not have an at-
titude he appears to be expressing, especially if the reason is mutually
believed, is a reason to think S is being nonliteral or that the CP is not in
effect. However, there are other possibilities. For instance, to avoid
admitting something or committing himself, S has good reason to ex.:.
press attitudes he does not have, despite its being obvious that he
doesn't have them. His insincerity is transparent, and yet by his utter-
ance he has provided H with a basis for determining precisely what S is
being insincere about. The obviousness of S' s insincerity does not pre-
vent him from performing the illocutionary act of expressing attitudes
he doesn't have. Rather, it prevents his utterance from providing (and
being R-intended to provide) sufficient reason for H to think he has
those attitudes. However, this does not mean his utterance provides no
reason, for unless it provided some reason by being R-intended to, it
would not be a case of obvious insincerity.
Obvious insincerity is not the only way in which S's utterance can
fail to be R-intended to provide sufficient reason for the hearer to
ascribe certain attitudes to the speaker. Another route is obvious su-
perfluity, where it is already mutually believed what S's attitudes are,
or where it is already mutually believed that, for example, H believes
what S believes or will do what S wants him to do. Here the reasons
that S's utterance would normally provide for attributing beliefs or
intentions to S do not need to be provided. But that does not mean that
they are not provided anyway. Of course, the hearer, in identifying the
expressed attitudes, would need to figure out why the speaker is
bothering to express them.
We have not attempted to enumerate the sorts of reasons a speaker
might have for expressing attitudes he obviously does not have. We
have pointed out only that he can successfully, however infelicitously,
perform the communicative illocutionary acts of expressing such at-
titudes. Why a speaker should do that is the hearer's problem, a prob-
lem that can arise only if the speaker is actually expressing certain
attitudes that he could not possibly have.
 Illocutionary Acts 61

Chapter Four Elaborating the Schema: to another member H, S is doing so with some recognizable illocution-
Illocutionary Acts ary intent.
Presumption of Literalness (PL): The mutual belief in the linguistic
community CL to the effect that if in uttering e, S could (under the
circumstances) be speaking literally, then S is speaking literally.
Then, S is F-ing that p if in the presence of some H, S utters some e in
language L, intending, and expecting (pursuant to the LP, the CP, and
the PL) H to recognize that he intends, H to infer (from the fact that S
means ... bye and the fact that S is thereby saying that *( ...p ... that S
is F-ing that p. That is, S intends, and expects H to recognize that he
intends, H to reason as follows:
Having surveyed the kinds of communicative illocutionary acts, we
will now refine the speech act schema by spelling out the ways in which Basis
illocutionary acts can be performed via the SAS. We will start with the Ll. S is uttering e. hearing Sutter e
most straightforward kind of case, the literal and direct act, then move L2. S means ... bye. Ll, LP, MCBs
to more complicated cases, elaborating the schema as we go and de- L3. S is saying that *( ...p .. . ). L2, LP, MCBs
lineating its relation to other aspects of the communication situation. L4. S, if speaking literally, is
F*-ing that p. L3, CP, MCBs
4.1. LITERAL (AND DIRECT) ILLOCUTIONARY ACTS L5. S could be F* -ing that p. L4, MCBs
L6. S is F*-ing that p. L5, PL
If the suggestions made so far are correct, we should be able to select
Consider a particular case. Let e be the sentence" John will pay Sam
appropriate values for e, p, and so on, specify the mutual contextual
back." H is to infer:
beliefs, and instantiate the SAS with examples of various types of
illocutionary acts, thereby characterizing their literal (and direct) per- Basis
formance. How does the SAS characterize the performance of literal Ll. S is uttering e. hearing Sutter e
(and direct) communicative illocutionary acts? Let's start with cases L2. S means 'John will repay
where the act is performed directly rather than by means of another Sam' bye. Ll, LP, MCBs
illocutionary act (that is, indirectly). For convenience we repeat the L3. S is saying that John will
presumptions introduced in chapter 1 which, together with MCBs, repay Sam. L2, LP, MCBs
sanction various steps of the SAS. L4. S, if speaking literally, is
constating (or ... ) that John
Linguistic Presumption (LP): The mutual belief in the linguistic com-
will repay Sam. L3, CP, MCBs
munity CL to the effect that:
L5. S could be speaking literally. L4, MCBs
i. the members of CL share L, and
'L6. (a) Sis constating (or ... ) that
ii. whenever any member S utters any e in L to any other member H,
John will repay Sam. L5,PL
H can identify what S is saying, given that H knows the meanings of e
(b) S is predicting that John
in L and is aware of appropriate background information.
will repay Sam. L6(a), MCBs
Communicative Presumption (CP): The mutual belief in the linguistic
Given that H has inferred L4 from L3, why does he infer L5, and how
community CL to the effect that whenever a member S says something
does he arrive at L6(b)?
The Theory 62
First, on what basis does H suppose S could be speaking literally?
That is, why does he infer L5? It would seem that the reasonableness of
an expectation that one's intentions will be recognized depends on
general coherence of belief as well as contextual and conversational
appropriateness, as these are judged by the speaker and the hearer.
Indeed, what we are working with are factors believed by the speaker
to be believed by the hearer to be coherent or appropriate, and vice
versa. The obvious problem is that what one person may be supposed
to believe is coherent or appropriate can differ wildly from what an-
other person may be supposed to believe. For instance, suppose S says
something that if taken literally in L is outrageously false or inappropri-
ate. Should H conclude that S does not mean bye what e means in L,
that S is not speaking literally, or that S holds some outrageous belief
(about the world at large or about the conversation in particular)?
We see no general way of deciding such a question. There would
seem to be no limit on the things one person can suppose another to
believe he believes-S can suppose H thinks S is a complete fool. The
best we can hope for is solutions by cases. It depends on the case what
is critical if Sand H are to mutually believe that they share (1) similar
idiolects, (2) similar general beliefs about the world, and (3) similar
conceptions of the nature, stage, and direction of the current talk-
exchange. We dealt with (1) in chapter 2, and because (2) is much too
general to handle here, let us focus on issues surrounding (3).
Without a shared conception of the nature, stage, and direction of the
talk-exchange, H could hardly tell whether S meant what he said. What
is said may well be in and of itself perfectly reasonable but con-
versationally inappropriate if construed literally or as S's complete
contribution to the talk-exchange at that point. In the course of a con-
versation on people's responsibilities to others, S might intone "No
man is an island, you know." Had H been contending that no one has
any connections with anyone else, S's remark could be interpreted as
an objection, as well as a statement, to the effect that one is indeed
affected by and responsible for others. It would be taken differently in a
conversation concerning the analytical properties of "man." Yet in
each case the sentence might well be meant literally, and it is not
outrageously false (it is obviously true). What, then, makes a contribu-
tion conversationally appropriate?
Following Grice (1975) we assume that cooperative conversations
are governed by certain maxims,! or as we prefer to call them because
they are defeasible mutual contextual beliefs, conversational presump-
Illocutionary Acts 63
tions. When a person fails to fulfill one of them, H will take S as having
spoken contextually inappropriately until or unless H finds a suitable
explanation to the contrary.
In the course of a talk-exchange speaker and hearer presume that at
any point in the talk-exchange,
Relevance (RE): The speaker's contribution is relevant to the talk-
exchange at that point.
This is a very general and powerful presumption; it is also vague. Just
what counts as being relevant? We do not have a general answer to this
question, 2 but there are two major parameters of relevance-force
and content. The following presumptions falling under (RE) pertain to
force.
Sequencing (SE): The speaker's contribution is of an illocutionary type
appropriate to that stage of the talk-exchange.
Sincerity (SI): The speaker's contribution to the talk-exchange is sin-
cere-the speaker has the attitudes he expresses.
Examples of compliance with these presumptions are obvious. Ceteris
paribus, questions are to be answered, requests and commitments ac-
knowledged, greetings reciprocated, constatives concurred with (or
dissented from, or elaborated upon), and so on. And all are to be done
sincerely where sincerity is possible. 3
Given that a certain type of illocutionary act is appropriate to the
present stage of a talk-exchange, there are presumptions relevant to
content. These fall into two major categories-quantity of information
and quality of information. Like Grice, we can begin by formulating the
presumptions for constatives:
Quantity (QT): The speaker's constative provides (or S assumes in
constating) just the requisite amount of information-not too much,
not too little.
Quality (QL):
i. The speaker attempts to make his constative true.
ii. The speaker has adequate evidence for what he constates (or as-
sumes in constating).
Analogs of these presumptions exist for other types of speech acts. For
directives, the presumptions are:
 The Theory 64
Quantity (QT): The speaker's directive provides (or S assumes in di-
recting) requisite information for compliance.
Quality (QL):
i. The speaker attempts to make his directive such that compliance is
possible. ("Don't ask for the Moon.")
ii. The speaker has reasons for what he directs (or assumes in direct-
ing),4 ("Leave." "Why?" "I don't know.")
For commissives and for acknowledgments there seems to be no pre-
sumption for quantity, only for quality:
Quality (QL): The speaker commits himself only to something he be-
lieves he is able to do.
Quality (QL): The speaker acknowledges only that which he believes to
have occurred.
Other presumptions concern not force or content but the manner of
performance, the way what is said is in fact said. Following Grice
(1975, 46) we will formulate them thus:
Manner (MA): The speaker speaks perspicuously, that is, S:
i. avoids ambiguity,
ii. avoids obscurity of expression,
iii. avoids unnecessary prolixity,
iv. is orderly.
Finally, there might be presumptions having nothing directly to do with
relevance but, rather, with the speech act as a social act. Two dimen-
sions that come to mind are politeness and morality. Because discuss-
ing these notions would take us far afield, we offer the following merely
as first approximations:
Politeness (PO): The speaker (in speaking) behaves politely, that is, S is
not offensive, abusive, rude, vulgar. 5
Morality (MO): The speaker (in speaking) behaves morally, that is, S:
i. does not reveal information he ought not reveal,
ii. does not ask for information he shouldn't have,
iii. does not direct H to do/tell something H shouldn't do/tell,
iv. does not commit himself to do something for H that H does not
want done.
Illocutionary Acts 65
These presumptions, from Quantity through Morality, need consider-
able refinement and supplementation (see Kempson 1975, chs. 7,8;
Hamish 1976b, 340-348). And there is the problem of how conflict
between presumptions is resolved. Nevertheless, we will press these
conversational presumptions into service in their present state.
We are now in a better position to give the notion of conversational
appropriateness some content. A speaker's contribution to the talk-
exchange is conversationally appropriate if and only if it accords with
those conversational presumptions in effect at that time. Then, other
things being equal, H will infer line L5 of the schema if he takes S's
F*-ing to be conversationally appropriate.
Our second question about inferring literal illocutionary acts was
how H infers L6(b) from L6(a). This is the problem of determining
which of the possible illocutionary acts that S might be performing
literally is the one S actually is performing. How does H recognize S's
specific illocutionary intent? If the CP is in effect, S will be presumed to
be speaking with recognizable illocutionary intent. If conversational
presumptions are in effect, H will conclude that S is speaking with the
illocutionary intent complying with them. In our original example, if it
is mutually believed that S can or will have no influence on John, then
S's saying that it will be the case that John repays Sam may be taken as
a prediction but not as a guarantee. If no one has claimed that John will
not repay Sam, then S's remark will not be taken as a dissentive. In this
way, specific MCBs and the relevant conversational presumptions
interact with what has been said to lead H to S's specific illocutionary
intent.
Three caveats: first, there is more to be said about how specific
illocutionary intent is identified, but, second, it should be realized that
often one speaks with no more specific illocutionary intent than, say,
just constating that something is (or will be) the case. Third, it is always
possible for H's identification of that intent to be vague, inaccurate, or
just plain wrong. After all, he can follow the SAS without filling it in
correctly.
4.2. NONLITERAL ILLOCUTIONARY ACTS
In circumstances where it would not be reasonable to suppose that the
speaker is F*-ing that p in saying that *( ...p ... ), the hearer will seek
(and will be intended to seek) an alternative explanation for S's utter-
The Theory 66
ance. That is, H will suppose that some other illocutionary act is being
performed which is not delimited by what S said and so is not being
performed literally. If the CP is in effect, H must be reasonably ex-
pected to be able to recognize that S intends to be F-ing thatP, and this
requires some sort of recognizable connection between what S said and
what S was intending to do. Accordingly, we can supplement the
schema as developed in section 4.1 to handle direct but nonliteral utter-
ances:
Basis
L5'. S could not (under the
circumstances) be F*-ing
thatp.
L4, MCBs
L6'. Under the circumstances
there is a certain recog-
nizable relation R between
saying that *( ...p ... ) and
some F-ing that P, such that
S could be F-ing that P. L3, L5', CP
L7. S is F-ing that P. L6', MCBs
A nonliteral illocutionary act has three basic ingredients: what is said,
what is done, and the (intended) relation between them. The schema
assigns two important roles to the MCBs: in the first step they signal
nonliterality; in the second step they guide H's search for S's nonliteral
illocutionary intent.
What are some of the ways utterances and MCBs can help signal
nonliteral acts? Although nonliterality can affect force as well as
content, usually only content is affected. For instance, consider exag-
geration, either understatement or overstatement (hyperbole). In un-
derstatement, one purports to claim what is in fact less, so to speak,
than one intends to claim. Since understatements are true if what is
intended to be communicated is true, recognition of S's communicative
intention cannot depend on saying something S obviously does not
believe, and so they do not violate the conversational presumption of
sincerity-one has told the truth. 6 Rather, one has not told the whole
truth and so has violated the presumption of quantity, as in:
(1) a. Not bad! (Very good! Great!)
b. It's OK. (Good!)
c. He's getting by. (He's doing fine.)
Illocutionary Acts 67
d. I wasn't born yesterday. (I'm not that naive.)
e. Boys will be boys.
f. No man is an island.
In overstatement, one purports to claim what is in fact more than one
intends to claim, as in:
(2) a. No one understands me. (Not enough people understand me.)
b. A pig wouldn't eat this food. (A person wouldn't eat it-ifhe had
a choice.)
c. Her eyes opened as wide as saucers. (Her eyes opened very
wide.)
d. I can't make a shot today. (I'm making very few.)
Since superlatives are a standard device for expressing extreme evalu-
ations, many overstatements use superlatives:
(3) That was the worst food I've ever had. (It was very bad.)
Advertisers make systematic use of overstatement in slogans:
(4) a. Paul Newman is Jesse James. (Paul Newman plays the part con-
vincingly, or with conviction.)
b. We do it all for you. (We look after your interests; you need do
nothing.)
c. When you say "Bud," you've said it all. (all that needs to be said
about beer)
d. If it's not Schlitz, it's not beer. (not the way beer should be)
e. The future is now. (You should prepare now for the future.)
In many of these cases there is some dimension toward one extreme of
which the claim would fall if literal, but literally the claim is false and is
contextually obviously so. Therefore, what the speaker is taken to be
doing is some act falling closer to the midpoint of that dimension.
If exaggeration is carried all the way along the dimension that relates
what is said to what is done, we get sarcasm, irony, and facetiousness.
In these cases one means roughly the opposite of what one says:
(5) a. Boy, this food is terrific! (terrible)
b. That argument is a real winner. (loser)
Although these acts can be nonliteral, it might be that an ironic or
sarcastic intonation contour signals one's intent. If so, such a contour
might be considered a negation operator converting the sentence into
The Theory 68 Illocutionary Acts 69

one for performing the act literally-just as rising intonation can con-
vert a declarative sentence into an interrogative.
Finally, there is the very difficult class of cases encompassed by the
term metaphor. Some metaphors are dead ("The leg of the table"),
others only moribund ("Are you a mouse or a man?" "Toss me some
cheese and you'll find out"). Literary critics have developed a rich,
though somewhat confusing, taxonomy of metaphors based largely on
the type of connection between what is said and what is meant. For
example, if the relation is part to whole it is called synecdoche:
(6) I've got three hands (workers) here to help.
If one thing bears a very close association to another, the expression is
sometimes classified as metonymy:
(7) a. The White House (the president or staff) said so.
b. The Crown (the monarch or staff) said so.
c. I have read all of Chomsky (Chomsky's works).
It would be a mistake to assume that all nonliteral uses of language, or
even just all metaphor, must be indicated by any single sort of cue or
violation of presumptions. In our examples so far, nonliterality has
been signaled in at least four ways:
Contradiction or anomaly: The future is now.
Conceptual truth: No man is an island.
Obvious factual falsehood: She's a gazelle.
Obvious factual truth: I wasn't born yesterday.
In the first two cases conceptual knowledge, plus context, is sufficient
to trigger completion of the nonliteral strategy. In the remaining two
cases H must identify what is being referred to, as well as some general
(but nonessential) properties held true of these referents, in order to
follow the nonliteral strategy. For example, H must suppose that "she"
in "She's a gazelle" is being used to refer to a certain female person and
that gazelles are graceful.
What sorts of relations to what is said can guide H to S's nonliteral
intent? We have seen three in operation, roughly as follows:
(Rl) Sarcasm, Irony: the opposite of what is said.
(R2) Figure of speech: a figurative or metaphorical connection.
(R3) Exaggeration: the next evaluation toward the midpoint of the rel-
evant scale.
Does H just flounder around with Rl-R3 in seeking S's illocutionary
intent, or are these relations ordered in some way? Interestingly, when
two are used in a single sentence at the same time, the effects can be the
same regardless of order. Consider a simple example. Mter-a gourmet
meal (mutually believed to be such) S utters "That was the worst
dinner in my life." H could infer either (8) or (9) with the same result:
(8) a. That was the best dinner in S's life. (by Rl)
b. That was a very good dinner. (from (8a) by R3)
(9) a. That was a very bad dinner. (by R3)
b. That was a very good dinner. (from (9a) by Rl)
The same can happen with some metaphors. Suppose that S utters
"Hey, that's a hot car!" referring to a stock Edsel. H could infer (l0) or
(11):
(10) a. That is a cold car. (by Rl)
b. That is a slow car. (from (lOa) by R2)
(11) a. That is afast car. (by R2)
b. That is a slow car. (from (lla) by Rl)
Is there an example that uses R2 and R3? One can exaggerate a
metaphor: S utters "She's a gazelle today," referring to someone who
moves none too gracefully on the court, but is doing better today. H
can infer:
(12) a. She is moving most gracefully today. (by R2)
b. She is moving very well (better) today. (from (l2a) by R3)
But these inferences cannot easily be reversed because e is already a
metaphor before exaggeration. Is it possible to make a metaphor out of
an exaggeration? Perhaps, but we have not found a plausible example.
If it is not plausible for R3 to apply before R2, the relations are
indeed ordered as they are numbered, since Rl can always apply first,
and R2 must come before R3. Thus H does not flounder around looking
for S's nonliteral illocutionary intent but rather has at least three in-
terpretive ubstrategies, which are utilized (tested, rejected, used) in
the order Rl, R2, R3. 7 Though these relations may generate the
metaphor-potential of various expressions, they do not select from the
possibilities they generate. For instance R2 does not itself determine
whether "Two hands showed up" is to be interpreted as 'two arms,'
'two pairs of arms,' or 'two workers.' To cut down these possibilities,
The Theory 70
MCBs must be invoked for communication to succeed. If context is
such that the sentence is understood as short for "Two hands showed
up for work this morning," then H will infer (and be intended by S to
infer) that two workers showed up. The same strategy can be applied to
the example "She's a gazelle today." Gazelles have many distinctive
properties, but only a few could reasonably be expected to be attrib-
uted to a tennis player (jumping ability, speed, grace) and in the context
of utterance these will either be further reduced by the conversational
presumptions (she has just jumped high for an overhead, and S may be
presumed to be commenting on that event) or the exact interpretation
will be left open. If left open, H will have only a general impression of
what S is intending to convey.
4.3. INDIRECT ILLOCUTIONARY ACTS
Though speaking literally in F*-ing that p by saying that *( ...p ... ), a
person can be indirectly performing another illocutionary act as well,
as illustrated by the utterances "The door is over there" and "My
mouth is parched." In the first case, S can be requesting H to leave
while stating that the door is over there; in the second, S can be re-
questing something to drink while informing H that his mouth is
parched. We will use the label indirect illocutionary act for an illocu-
tionary act that is performed subordinately to another (usually literal)
illocutionary act. It is indirect in the sense that its success is tied to the
success of the first act. That is, securing uptake requires H to identify
the indirect act by way of identifying the first act.
This characterization of an indirect act is similar to that of Searle
(1975a, 59-60) but differs from Searle's in covering nonliteral as well
as literal cases. It differs substantially from the characterization made
by some linguists. For them a disparity between the surface form of a
sentence and the illocutionary act performed in its utterance is critical
for the act's being indirect. For instance, Sadock (1974, 73) writes,
"Based on this discrepancy between surface form and use, such sen-
tences have been termed indirect illocutions" (see also Davison 1975,
143-144; Herringer 1972, ch. 3). On this use an illocutionary act per-
formed in uttering e is indirect jus t in case the type of illocutionary act
associated with the surface form of e is distinct from the type of act
performed. Notice that these two kinds of characterization are impor-
tantly different. If a surface interrogativ~, say, is given the semantics of
Illocutionary Acts 71
an imperative by s9me grammar, then on the linguistic characteriza~ion
the request made by its utterance is indirect, but on our charactenza-
tion it would be direct. In addition, the linguistic characterization
would not represent as indirect an act of the same force but with differ-
ent content from the act performed directly.
Let us supplement the basic SAS to cover literally based indirect
acts:
Basis
L 7'. S could not be merely
F*-ing that p. L6, MCBs
L8. There is some F -ing that P
connected in a way identi-
fiable under the circum-
stances to F*-ing thatp, such
that in F* -ing that p, S
could also be F-ing that P. L7', CP
L9. S is F*-ing that p and
thereby F-ing that P. L8, MCBs
Note that line L9 has S both F*-ing that p and F-ing that P. Whereas
the former is literal and direct, the latter can be either literal or non-
literal. In most cases the indirect act will be nonliteral, such as when
one requests a drink by stating that one's mouth is parched. However,
it is possible for the indirect act to be literal as well, though clear cases
are hard to motivate. Consider the case of warning H that the bull is
about to charge. Both warning (in this example) and stating are con-
stative and so are L-compatible with e, and the propositional content is
the same. Thus both acts are literal. To be indirect, and not just simul-
taneous, the warning must be performed by means of the stating. In this
case it could be accomplished by way of an MCB that bulls are danger-
ous.
Not all indirect acts need be literally based. It is possible to speak
indirectly by speaking nonliterally - for instance, sarcastically. A
mother might say (sarcastically) to her son "I'm sure the cat likes
having its tail pulled," intending to (a) directly and nonliterally state
that S is sure that the cat does not like having its tail pulled, 8 and (b)
indirectly request that H stop pulling the cat's tail. Accordingly, we
extend the SAS to nonliterally based indirect acts:
 r
:
The Theory 72
Basis
L8'. S could not merely be F-ing
that P. L7, MCBs
L9'. There is some F' -ing that Q
connected in a way identi-
fiable under the circum-
stances to F-ing that P,
such that in F -ing that P, S
could also be F'-ing that Q. L8', CP
LI0. S is F-ing that P and
thereby F'-ing that Q. L9', MCBs
Notice that with respect to indirect acts, whether literally or nonliter-
ally based, the MCBs play two characteristic roles. First (at line L7' or
L8') they help determine that the direct illocutionary act cannot rea-
sonably be taken to be the sole act being performed by S in his utter-
ance. Then (at L8 or L9') the MCBs contribute to determining the
identity of the indirect act.
We have already seen how the conversational presumptions initiate
nonliteral strategies; how do they initiate indirect strategies? The main
cue in these cases involves the idea of S's making an adequate con-
tribution to the talk-exchange at that point. H is to reason that S's
direct contribution to the talk-exchange cannot (pursuant to the con-
versational presumptions) be S's total contribution at that point be-
cause it is inadequate in some recognizable respect. That is, it violates
some conversational presumption. Saying this, though, does not de-
lineate the contributions of these presumptions to the process leading
H to S's indirect illocutionary intent.
What then is the process leading H from S's direct to S's indirect
illocutionary intent? What sorts of connections can S expect H to
utilize? Consider again some of the examples we have mentioned so
far:
(13) a. The door is over there. (used to request someone to leave)
b. My mouth is parched. (used to request a drink)
c. I'm sure the cat likes having its tail pulled. (used to request H to
stop pulling the cat's tail)
To these cases we can add cases like the following:
(13) d. You're the boss. (used to agree to do what H says)
e. I should never have done that. (used to apologize)
Illocutionary Acts 73
f. Did you bring any tennis balls? (used to inform H that S did not
bring any)
g. It's getting late. (used to request H to hurry)
The schema at lines L8 or L9' reflects the fact that the connection
between the direct and indirect intent can be extremely context depen-
dent, there being few substantive generalizations spanning all cases.
This can be better appreciated after we investigate how the inference
might go in some of the examples in (13).
(13a) Suppose S utters "The door is over there" to H, thereby stating
that the door is over there. Then H may infer:
L7' . S could not be merely stating that the door is over there.
Basis: It is MB-ed that Sand H are having a dispute and the
location of the door is irrelevant to the discussion so far.
L8. There is some F-ing that P connected in a way identifiable under
the circumstances to stating that the door is over there, such that
in stating that the door is over there S could also be F-ing that P.
Basis: L7', CPo
L9. S is stating that the door is over there and thereby requesting H to
leave.
Basis: It is MB-ed that doors are customary means of leaving a
room. S's most obvious additional illocutionary intent under these
circumstances (a dispute) is to request H to leave.
(13b) Suppose S utters "My mouth is parched" to H, thereby stating
that S's :r;nouth is very dry. Then H may infer:
L 7'. S could not be merely stating that S's mouth is very dry.
Basis: It is MB-ed that S had earlier come in out of the sun, this
was not the subject of previous conversation, and so it is irrele-
vant.
L8. There is some F-ing that P connected in a way identifiable under
the circumstances to stating that one's mouth is very dry, such that
in stating that his mouth is very dry S could also be F-ing that P.
Basis: L7', CPo
L9. S is stating that S's mouth is very dry and thereby requesting a
drink.
Basis: It is MB-ed that people whose mouths are very dry are
thirsty, that being thirsty is uncomfortable, and that drinking
something relieves thirst. S's most obvious additional illocution-
ary intent under the circumstances (S's thirst) is to request a
drink.
The Theory 74
(13c) Suppose S utters "I'm sure the cat likes having its tail pulled"
to H, thereby (nonliterally) claiming that S is sure that the cat does not
like having its tail pulled. H may infer:
L8'. S could not be merely claiming that the cat does not like having its
tail pulled.
Basis: It is MB-ed by Sand H already that this is true, so S would
violate the presumption of quantity.
L9'. There is some F*-ing that Q connected in a way identifiable under
the circumstances to claiming that the cat does not like having its
tail pulled, such that in claiming that the cat does not like having
its tail pulled S could also be F*-ing that Q.
Basis: L8', CP.
LIO. S is claiming that the cat does not like having its tail pulled and
thereby requesting H to stop it.
Basis: It is MB-ed that H's pulling the cat's tail hurts the cat and
that S cares about the cat. S's most obvious additional illocution-
ary intent under these circumstances is to request H to stop pull-
ing the cat's tail.
(13d) Suppose S utters "You're the boss" to H, thereby asserting
that H is the boss. H may infer:
L 7'. S could not be merely asserting that H is the boss.
Basis: It is MB-ed by Sand H that this is true, so S would violate
the presumption of quantity.
L8. There is some F-ing that P connected in a way identifiable under
the circumstances to asserting that H is the boss, such that in
asserting that H is the boss S could also be F -ing that P.
Basis: L7', CPo
L9. S is claiming that H is the boss and thereby agreeing to do what H
says.
Basis: It is MB-ed that H wants S to do A, that S is reluctant to do
so, but that because of H's position of authority over S, S has to
do A. S's most obvious additional illocutionary intent under these
circumstances is to agree to do A.
(13e) Suppose S utters "I should never have done that" to H,
thereby stating that S should never have done some specific act A. H
may infer:
L7'. S could not be merely stating that S should never have done A.
Basis: It is MB-ed by Sand H that A adversely affected H, that S
Illocutionary Acts 75
should not do such things, etc. So S would violate the presump-
tion of quantity.
L8. There is some F-ing that P connected in a way identifiable under
the circumstances to stating that one should never have done A,
such that in stating that one should never have done A one could
also be F-ing that P.
Basis: L7', CPo
L9. S is stating that S should never have done A and thereby apol-
ogizing for having done A.
Basis: It is MB-ed that people often regret doing things they be-
lieve they should not have done. S's most obvious additional
illocutionary intent under these circumstances is to apologize
for doing A.
(13f) Suppose S utters "Did you bring any tennis balls? to H, thereby
asking H whether H brought any tennis balls. H may infer:
L7'. S could not be merely asking whether H brought any tennis balls.
Basis: It is MB-ed that it is S's tum to supply the balls, so a mere
question would be irrelevant.
L8. There is some F-ing that P connected in a way identifiable under
the circumstances to asking whether H brought any tennis balls,
such that in asking whether H brought any tennis balls S could
also be F -ing that P.
Basis: L7', CPo
L9. S is asking whether H brought any tennis balls and thereby claim-
ing that he (S) did not bring any.
Basis: It is MB-ed that it is S's tum to supply the balls, so S would
not need to ask unless he forgot. S's most obvious additional
illocutionary intent under these circumstances is to claim that S
did not bring any balls.
As these examples illustrate, there is a heterogeneous variety of
connections between S's direct illocutionary intent and S's indirect
illocutionary intent. However, some connections are more regular and
systematic than others.
(14) a. I want you to move over. (used to request H to move over)
b. I intend to be there. (used to promise to be there)
C. I regret having done that. (used to apologize)
In these cases S directly states that he has an attitude which is ex-
The Theory 76 Illocutionary Acts 77

pressed in the performance of the indirect act. That this connection is And possibly (literally based indirect),
quite straightforward and recognizable may account for the systematic L 7'. S could not be merely
nature of these examples, as well as of the following: F*-ing that p. L6, MCBs
L8. There is some F-ing that p
(15) a. Are you going to put out the garbage? (used to request H to put
connected in a way identi-
out the garbage)
fiable under the circum-
b. You won't come back here again. (used to order H not to re-
stances to F*-ing that p,
turn)
such that in F*-ing thatp, S
In these cases S directly expresses desire to be informed about, or could also be F-ing that P. L7 ' , CP
belief in, the act indirectly requested. L9. S is F*-ing that p and
thereby F-ing that P. L8, MCBs
(16) a. Ought you to smoke here? (used to request H to stop smoking)
b. You shouldn't smoke here. (used to request H to stop smoking) Or (direct nonliteral) ,
L5 ' . S could not (under the
In these cases S directly expresses either desire to be informed about, circumstances) be F*-ing
or belief in, a certain reason for H's action. Such obvious connections thatp. L4, MCBs
as those illustrated in (14) -(16) can give rise to standardized illocu- L6 ' . Under the circumstances
tions, whereby segments of the SAS are short-circuited. This interest- there is a certain recog-
ing phenomenon is explained in chapter 9. nizable relation R between
saying that p and some
4.4. THE ELABORATED SCHEMA F-ing that P, such that S
could be F-ing that P. L3, L5 ' , CP
We can now pull together the elaborated version of the SAS that covers L7. S is F-ing that P. L6 ' , MCBs
nonliteral and indirect, as well as literal and direct illocutionary acts. S And possibly (nonliterally based indirect),
is F-ing that P if in the presence of some H, S utters some e in some L8 ' . S could not merely be F-ing
language L intending, and expecting (pursuant to the LP, the CP, and that P. L 7, MCBs
the PL) H to recognize that he intends, H to infer (from the fact that S L9 ' . There is some F' -ing that Q
means ... bye and the fact that S is thereby saying that *( ...p ... that connected in a way identi-
S is F-ing that P. On occasion S may be also pl-ing that Q. That is, S fiable under the circum-
intends, and expects H to recognize that S intends, H to reason thus: stances to F-ing that P,
Basis such that in F-ing that P, S
L 1. S is uttering e. hearing Sutter e could also be F'-ing that Q. L8 ' , CP
L2. S means ... bye. Ll, LP, MCBs LI0. S is F-ing that P and
L3. S is saying that *( ...p ... ). L2, LP, MCBs thereby F'-ing that Q. L9', MCBs
L4. S, if speaking literally, is According to the elaborated schema, H's identification of an il~ocu
F*-ing that p. L3, CP, MCBs tionary act is the result of applying a selection of inference strategles to
Either (direct literal), the utterance (assuming H has inferred Ll):
L5. S could be F*-ing that p. L4, MCBs
Locutionary Strategy (LS): Given Ll, infer L2, L3, L4.
L6. S is F*-ing that p. L5,PL
Direct Literal Strategy (DLS): Given L4 (from LS), infer L5, L6.
The Theory 78 Illocutionary Acts 79

(Ll) Utterance of e Literal strategy:

There is nothing (under the circumstances) incompatible with the

$
(L4) F*-ing that p,
supposition that S is telling me (H) that S's mouth is very dry. So
L5. S could be telling me (H) that S's mouth is very dry. So
L6. S is telling me (H) that S's mouth is very dry.
if literal
I Literally-based indirect strategy:
(a) I (H) have not asked for or otherwise indicated in any way a

(L6) F*-ing that p

$
(L7) F-ing that P
desire for this information about S's mouth, nor has this come up
in the conversation. So
(b) If L6 is all S is doing, S would be being irrelevant (violating

~
presumption RE).

(L9) F-ing that P

$
(LIO) F' -ing that Q
Figure 4.1 Organization of strategies
Literally-based Indirect Strategy (LIS): Given L6 (from DLS), infer
LT, LS, L9.
Direct Nonliteral Strategy (DNS): Given L4 (from LS), infer L5', L6',
L7.
Nonliterally-based Indirect Strategy (NIS): Given L7 (from DNS), infer
LS', L9', LI0.
All this is represented diagrammatically in figure 4.1.
We can now use an example presented earlier to show in more detail
how to instantiate the SAS. Imagine the following situation: S has been
working in the hot sun, H is in the kitchen, H asks S how the work is
going, and S says that it is going fine and then adds "My mouth is
parched." We will assume that these facts are among the MCBs held by
Sand H and that the LP, the CP, the PL, and the conversational
presumptions are in effect. Then S intends (and expects H to recognize
that S intends) H to reason thus:
Locutionary strategy:
Ll. S is uttering "My mouth is parched."
L2. S means 'My mouth is very dry' by "My mouth is parched."
L3. S is saying that S's mouth is very dry.
L4. S, if speaking literally, is telling me (H) that S's mouth is very dry.
(c) But presumably S is being relevant. So
L7'. S could not be merely telling me (H) that S's mouth is very dry.
So
LS. There is some further act connected in a way identifiable under the
circumstances to S's telling me that his mouth is very dry, such
that in telling me that his mouth is very dry, S could also be
performing that act.
(a) One's mouth being very dry is usually symptomatic of thirst.
(b) Being thirsty is a state one wants to relieve by drinking some-
thing. So
(c) S desires a drink.
(d) S intends me (H) to infer (c).
(e) The obvious explanation for (d) is that S wants me to satisfy
this desire, viz., by getting S a drink.
(f) S intends me (H) to infer (e).
(g) The obvious explanation for (f) is that S intends me (H) to get
S a drink because S desires me (H) to. So
(h) S is expressing the desire that I (H) get S a drink and the
intention that I (H) do so because S desires me to. So
L9. S is telling me (H) that S's mouth is very dry and is thereby asking
me (H) to get S a drink.
This sort of inference operates in a wide variety of other examples,
such as "Is your head cold, Private?" used to order someone to take off
his hat, "I have the car tonight" used to offer to give someone a ride
that night, and "That was a tough match" used to congratulate some-
one.
Do all indirect illocutionary acts involve such complex inferences?
This question stems from two complementary considerations. First,
The Theory 80
certain types of expressions are especially suited to the indirect per-
formance of certain types of illocutionary acts. Moreover, certain types
of illocutionary acts are especially suited to being indirectly performed
by the use of certain types of expressions. A good deal of linguistic
literature, as well as philosophical literature prompted by it, has fo-
cused on these points. Sadock (1974), for instance, thinks that sen-
tences like the following are ambiguous as to their force, with one
meaning being interrogative, the other "idiomatically" imperative:
(17) a. Can you pass the salt? (Pass the salt!)
b. Will you be quiet? (Be quiet!)
c. Why not (don't you) shut up? (Shut up!)
Searle (1975a), on the other hand, thinks that these sentences are not
ambiguous, but are "conventionally used" to request. We will post-
pone discussion of their positions (and of our alternative) until chapter
9, since almost everyone seems to agree that except for these special
cases, most indirect acts should be analyzed inferentially and conver-
sationally.
In summary, whether literal, nonliteral, or indirect, an illocutionary
act must be such that if it is to be performed successfully and felici-
tously, the speaker can reasonably expect it to be identified by the
hearer. This expectation is reasonable if it is based on what we have
called mutual contextual beliefs, together with the various presump-
tions. What constitutes a reasonable expectation is complicated by the
many sorts of inference routes that in different cases a hearer may be
intended and expected to follow to arrive at the identity of the illocu-
tionary act. We do not propose to catalog them here in any more detail
than represented by the SAS, but whatever the exact route in a par-
ticular case, presumably (that is, by the CP) it must be such that the
hearer can find it. Therefore, the utterance must be made with the in-
tention that H can find the inference route, and part of what the hearer
takes into account in trying to find the route is that the utterance is
made with that intention.
Detailed as it is, the SAS does not represent the precise form of
inference (to be) made by the hearer. Left open are the questions of just
which mutual contextual beliefs are activated and of just how they
enable the hearer to find the right candidate for the speaker's illocu-
tionary intent (not that there is ever any guarantee of success). A
complete account of the hearer's inference would require a systematic
theory of saliency or obvious obviousness, a notion introduced by
Illocutionary Acts 81
Schelling (1960, ch. 3) in connection with coordination problems. To
our knowledge no such theory has yet been developed. Until it has, the
SAS can represent only the general pattern of the hearer's inference.
4.5. PERLOCUTIONARY ACTS
In section 1.7 we restricted perlocutionary acts to the intentional pro-
duction of effects generated off of steps of the SAS, particularly off of
illocutionary acts. In chapter 3 we noted the perlocutionary intentions
correlated with the major types of illocutionary intents:
Constative: that H believe that P.
Directive: that H (intend to) do A.
Commissive: that H believe S intends to fulfill his obligation to do A.
Acknowledgment: that H believe S has the appropriate feeling.
In order to exclude irrelevant sorts of effects, such as neuron firings,
we restrict perlocutionary acts to the (intentional) production of certain
kinds of effects on H, namely, psychological states or intentional
actions.
Unlike illocutionary intentions, perlocutionary intentions need not
be recognized or intended to be recognized. They can be intended to be
recognized, although in some cases (such as misleading or impressing
someone) the hearer is intended not to recognize them. Not only can
perlocutionary intentions be overt, they can be reflexive; but even
when they are reflexive, they are not communicative in the way that
illocutionary intentions are, for their fulfillment consists not in their
recognition but in the production of some further effect.
The restrictions we have made on perlocutionary acts have at least
two beneficial results. First, we can account for many of the differences
(both well known and not so well known) between perlocutionary and
illocutionary acts. Second, by restricting perlocutionary acts to (in-
tended) effects generated off of steps of the SAS, the elaborated
schema allows us to systematize some complicated and subtle perlocu-
tionary actions and intents.
4.5.1. Perlocutions versus Dlocutions
That perlocutionary effects are produced off of the SAS helps to ex-
plain why illocutionary acts are characteristically means to perlocu-
The Theory 82 Illocutionary Acts 83

tionary ends, and not vice versa. And that illocutionary intentions do, make H mad (in stating that P) can itself contribute to making H mad.
but perlocutionary intentions do not, need to be recognized to be ful- Second, by F-ing that P, S may (intentionally) perform some perlocu-
filled helps to explain why there should be illocutionary (performative) tionary act and, in virtue of doing that, intend to produce some further
prefixes as in (18), but no perlocutionary prefixes as in (19). perlocutionary effect. For instance, in saying that the bull is about to
charge, S might predict that the bull will charge. By predicting this, S
(18) I (hereby) promise (order, state to, admit to, etc.) you that ... may convince H that it will charge and by convincing H of this, S may
(19) *1 (hereby) persuade (convince, surprise, etc.) you that ... frighten H enough to get H to move quickly out of danger. All of this
Furthermore, because there are degrees of conviction, motivation, in- could have been intended by S.
tention, and so on, perlocutionary intentions can have degrees of suc-
cess in a way that illocutionary intentions cannot. H can be partially or
completely convinced by S's statement in a way that S cannot partially
or completely make the statement. (Of course, H can be more or
less accurate in his identification of the contents of S's illocutionary
intention.)

4.5.2. Perlocutions and the SAS

Every major step in the schema can give rise to some distinct perlocu-
tionary effect and so be instrumental to the performance of some per-
locutionary act. Not only does the schema enable us to identify these
relationships systematically, it gains support from their very existence.
It is hard to imagine how steps in the schema could have psychological
effects without also having some psychological reality. Consider:
Ll. By uttering "Don't wake up," S awakened H.
L2. In uttering "He cleaves to the principle" and meaning 'He adheres
to the principle' by it, S impressed H with his vocabulary.
L3. By saying that religion is the opium of the people, S offended H.
L6. By predicting that Truman would win the election, S convinced H
that he was clairvoyant.
L7. By (directly and nonliterally) stating that H could not have it both
ways in uttering "You can't eat your cake and have it too," S
reminded H that he usually remembered this proverb backwards.
L9. By (directly) demeaning H's character and thereby (indirectly)
challenging H to fight, S both irritated H and amused H with his
pretension.
In addition to these sources of perlocutionary effect, there are some
partially perlocutionary sources as well. First, H's realization that S
has a particular perlocutionary intention can itself give rise to a per-
locutionary effect. For instance, the recognition of S's intention to
 Speech Act Schema 85

Chapter Five The Status and Scope of the attem of inference hearers actually make in identifying speakers' in-
Speech Act Schema P l
tentions. The claim that it does "
so is an extreme y strong empinca1
hypothesis, one for which we have provid~d little evidence, .and so we
must show why the SAS is at least a plausIble model of the Inferences
that hearers actually make. Otherwise, what reason is there for taking it
seriously? One possible response is that the SAS does not represent the
'nferences hearers actually make but constitutes rational reconstruction
~f such inferences. 2 Rather than weaken our claim for the SAS in this
way, we hope to show that the SAS is much more realistic psychologi-
cally than it might seem, given its ostensible complexity.

5.1. THE SAS AND THE ANALYSIS OF LINGUISTIC

Having elaborated the speech act schema we must now explain (1) its
COMMUNICATION
role in the analysis of linguistic communication, (2) the nature of the
inferences it represents, (3) why such inferences can plausibly be at-
Our first task is to indicate the relation between the SAS and our anal-
tributed to people, and (4) what the schema leaves out.
ysis of linguistic communication. In our account, li~guistic c?mm~ni
The SAS represents the pattern of inference a hearer follows in iden-
cation essentially involves issuing an utterance WIth an R-mtentlon
tifying a speaker's illocutionary intent. We have claimed that linguistic
whose fulfillment consists in its recognition. Our taxonomy of com-
communication essentially involves the speaker's issuing an utterance
municative illocutionary acts (in chapter 3) catalogs what sorts of
with an R-intention whose fulfillment consists in its recognition. Our
R-intentions are communicative. An R-intention whose fulfillment con-
analysis of linguistic communication does not require that this R-
sists in its recognition is an expression of attitude; it is fulfilled if the
intention include the intention that the hearer make a detailed inference
hearer identifies that attitude in the way intended.
in exactly the form of the SAS. Such a requirement would preclude our
The analysis of linguistic communication does not require that the
concept of linguistic communication from applying to the utterances of
speaker's R-intention include all the details represented in the SAS
everyday discourse. The SAS contains many steps as well as citations
about how his illocutionary intent is to be recognized. Then how should
of mutual contextual beliefs and the several presumptions; to require
we construe a case in which the speaker intends the hearer to infer his
the speaker to intend the hearer to make an inference of exactly the
intent in one specific way but the hearer identifies it (correctly) in some
form of the SAS in every detail would be tantamount to attributing to
other way? It might be argued that the speaker has not performed a
the general public our theory of understanding communicative illocu-
successful act of linguistic communication, even though the hearer
tionary acts, as encapsulated by the SAS. On the other hand, we can-
does identify his illocutionary intent. The underlying assumption here
not allow the speaker to have just any kind of intention as to how his
is that if there is a discrepancy between the inference H makes and the
illocutionary intent is to be recognized and his illocutionary act thereby
inference he is intended to make-even if he manages to identify S's
identified. Without intending the hearer to follow the SAS in detail, the
intent-then S has not succeeded in linguistically communicating his
speaker must at least R-intend the hearer to identify his illocutionary
intent to H. To be problematic, presumably the discrepancy in question
intent on the basis of what he utters, in accordance with the linguistic
must be not just one of degree (the hearer following the SAS without
presumption and the communicative presumption. That, we claim, is
being intended to follow its exact details) but one of outright conflict.
necessary and sufficient for attempted linguistic communication. 1
Does such a conflict generate a counterexample to our analysis of
If the full-blown SAS is not embodied in the (conceptual) analysis of
linguistic communication?
linguistic communication, what is its status? If the SAS is not part and
Suppose S says to H "Dinner is ready," thereby indirectly request-
parcel of speakers' communicative intentions, it can represent only the'
ing H to come to the table. S intends H to reason as follows:
The Theory 86
1. S is uttering "Dinner is ready."
2. By "Dinner is ready" S means 'Dinner is ready.'
3. S is saying that dinner is ready.
4. S is stating that dinner is ready.
5. S is not merely stating that dinner is ready.
6. H is hungry.
7. S wants H (and is requesting H) to come to the table.
However, after step 5 H reasons differently:
6'. S gets upset whenever I show no interest in S's cooking.
7. S wants me (and is requesting me) to come to the table.
This is a clear case where the hearer infers the speaker's illocutionary
intent but not as he was intended to infer it. Is it therefore not a case of
genuine communication, or is it a genuine case notwithstanding the
discrepancy in how the illocutionary intent is identified? We are in-
clined to say that it is a case of successful requesting and that the
discrepancy is not great enough to rule it out as linguistic communica-
tion. To be sure, there is an element of misunderstanding between S
and H, but not about S's illocutionary intent. H misunderstands S's
reason for having that intent and, derivatively, H's reason for identify-
ing it is different from the one S intended. Nevertheless, it seems to us
that S has succeeded in expressing his desire that H come to the table
and has thereby successfully performed the illocutionary act of re-
questing.
The discrepancy in this example concerns the relevant mutual con-
textual belief. Discrepancies regarding other inference elements in the
SAS also can occur without vitiating successful linguistic communica-
tion and the performance of an illocutionary act. A speaker can intend
to utter one thing, but utter another. He might utter "Foreman dealt
Frazier a blushing crow." The hearer, if he recognizes this spoonerism,
can correct for it and go on to make the intended inference. Similarly,
with a malapropism the hearer can again make the appropriate adjust-
ment. 3 Indeed a speaker, knowing that H misuses a certain word, can
intentionally misuse it, relying on H's mistaken belief about what the
word means; S might use "enervated" to mean 'energetic,' thereby
getting his misinformed audience to think that this is what he (S) thinks
it means. This is an example of how there can be a discrepancy be-
tween the inference S intends H to make and the inference H thinks he
is intended to make (and makes) without vitiating successful communi-
Speech Act Schema 87
cation. Furthermore, in such a case the speaker intends the hearer's
inference to contain a false step, namely, that "enervated' means
'energetic. '
In general, it seems to us that minor discrepancies between the infer-
ence intended and the inference made are compatible with successful
linguistic communication, provided of course that the hearer does
manage to identify the speaker's illocutionary intent. We do not see
how to demonstrate this, though obviously if successful communica-
tion required perfect congruence between the inference intended and
the inference made, successful communication would rarely take place.
Nevertheless, there is a very difficult question that a thoroughgoing
analysis would have to answer: what is the difference between a minor
discrepancy and a discrepancy large enough to vitiate communication,
even though the hearer does manage to identify the speaker's intent?
Lacking a systematic account of this difference, we can still say that in
everyday life minor discrepancies occur all the time without even being
noticed, much less without disrupting communication.
Notice further that although successful communication requires that
the hearer identify the speaker's illocutionary intent and recognize it as
intended to be recognized, it does not require that the speaker and the
hearer mutually believe that the hearer has identified the speaker's
illocutionary intent. Even when communication has succeeded, it is
possible for S to believe it has not succeeded (that H has misidentified
S's illocutionary intent) or to believe that H believes it has not suc-
ceeded. For that matter, H might believe, quite mistakenly, that he has
misidentified S's illocutionary intent. In practice, to the extent that the
success of communication is verified at all by Sand H (countless fail-
ures undoubtedly go unnoticed) verification is generally achieved not
by repetition, direct inquiry, hearer paraphrase, and the like, but by a
sense of coherence between the utterance in question and subsequent
ones.
Although successful communication does not require verification or
mutual belief, there seems to be (along with the communicative pre-
sumption) a presumption of understanding, of communicative success.
Just as there is a presumption that the speaker's illocutionary intent be
identifiable, so there is a presumption that the hearer has succeeded in
identifying it. If there is no indication to the contrary (a request for
repetition, paraphrase, or elaboration, or even a mere sign of bewilder-
ment), it is assumed that successful communication has been achieved.
In parti~ular, if there is no expressed doubt by H (or by S) concerning
The Theory 88
success, the conversation proceeds normally. Moreover, if there is
some discrepancy between the inference intended and the inference
made, it is highly unlikely, assuming the hearer has correctly identified
the speaker's illocutionary intent, that this discrepancy will become an
issue.
We have seen that minor discrepancies do not vitiate linguistic com-
munication. However, not just any hearer inference to the right illocu-
tionary intent counts as successful communication. The hearer might
totally misunderstand S and yet, quite coincidentally, identify S's illo-
cutionary intent anyway. Surely that would not count as successful
communication. But why not? We want to say that even when there are
discrepancies of the sorts we have mentioned, still the hearer must
make his inference on the basis of what the speaker utters (or intends to
utter) and thereby on what the speaker says (or intends to say).
Moreover, only if the communicative presumption applies to the con-
text of utterance can the hearer reasonably take the speaker to have a
certain illocutionary intent that he intends the hearer to recognize (on
the basis of what is uttered). The presence of the CP, together with the
LP, severely constrains the possible inferences the speaker can rea-
sonably expect, and thus intend, the hearer to make and, correlatively,
the possible inferences the hearer does make. The CP requires that the
illocutionary intent be identifiable on the basis of what is uttered, and
the CP is itself activated (if not activated already) by the very fact of
utterance. When an utterance is made under one of the special cir-
cumstances in which the CP is suspended (recitation, elocution lesson,
quotation), the speaker cannot reasonably expect the hearer even to
seek, much less identify, some illocutionary intent, and the hearer has
no reason to do so, unless there is some special indication. 4
We suggest, then, that communication has succeeded just in case the
hearer correctly identifies the speaker's illocutionary intent on the basis
of what the speaker utters, under the supposition (normally justified by
the CP) that this intent is intended to be recognized. It does not matter
precisely how identification is made. It does not even matter that the
speaker may intend the hearer to make a false step in his inference (and
that the hearer makes it), or that the hearer may attribute an intention
to the speaker by correcting for a slip of the tongue or a malapropism.
Part of the reason that such discrepancies do not vitiate linguistic com-
munication is that the existence of a communicative intention is deter-
mined independently of its identity. Its existence is determined in.
Speech Act Schema 89
virtue of the communicative presumption-what is in question for the
hearer when the CP is in effect is not the existence but the identity
of the speaker's communicative intention. s However, an account is
needed of just how big a discrepancy (and of what sorts) is compatible
with successful linguistic communication.
5.2. THE EXPLANATORY VALUE OF THE SAS
The speech act schema represents, we claim, the pattern of inference
hearers usually make in identifying speakers' communicative inten-
tions. This pattern of inference is complex, and to attribute it to people
in ordinary communication situations is to impute to them complex
cognitive abilities that are exercised whenever they are addressed. (We
have not attributed to speakers the intention that hearers identify
communicative intentions precisely in accordance with the SAS.)
In attributing this complex pattern of inference to people in their role
as hearers, we are offering a psychological description (or at least the
form of one) of what goes on in everyday communication. In the next
section we endeavor to show that this description is psychologically
realistic (we cannot prove that it actually applies). For the moment, we
wish to show what explanatory value it has, assuming it to be genuinely
applicable to the inferences that hearers make. Its explanatory value,
we suggest, is twofold. (1) From the point of view of the hearer, to go
through an inference in the pattern of the SAS is in effect to provide
himself with an explanation of the speaker's utterance: to explain S's
utterance is to identify the intention with which it is issued. (2) From
the standpoint of psychological explanation, the organization of in-
gredients in the SAS provides a framework in terms of which the ability
of hearers to identify speakers' illocutionary intents can be described
and ultimately explained.
What sort of explanation does the hearer seek of the speaker's utter-
ance? The utterance is the act of uttering a certain sentence with a
certain meaning, and the hearer's identification of it (up to ambiguity)
involves nothing more than the exercise of the hearer's linguistic com-
petence. However, he seeks also to identify the locutionary and the
illocutionary act performed by the speaker in his utterance, and this
involves ascribing intentions to the speaker, in particular, the intention
to be performing a certain illocutionary act (by way of performing a
certain locutionary act). To identify the intended illocutionary act is to
 The Theory 90
explain the utterance. The sort of explanation involved here is just
what we use whenever we explain what somebody is doing by citing the
intention with which he is doing it. 6
For psychology the organization of ingredients in the SAS provides
a framework in terms of which the abilities and performances of
speaker-hearers can be described in detail. Of course, the SAS does not
cover everything involved. There is the further problem for cognitive
(and social) psychology of determining how social norms, mutual con-
textual beliefs, and other beliefs about the social situation enter into
concrete communicative intentions and inferences. Moreover, the SAS
abstracts from the ongoing nature of verbal interchange. Sentences are
produced one after another, often in fits and starts, with each speaker-
hearer cognitively engaged not only in the production or perception of
strings of words but in the social interchange of the conversation itself,
as well as in private cognitive activity, which is not only unexpressed
but may even be irrelevant to the social situation. Producing a sentence
to express one's illocutionary intent requires selecting what to express
and finding the words to express it, all the while thinking on what the
conversation is about.
We have offered the SAS as representing inferences the hearer has to
make to identify the speaker's illocutionary intent. If hearers generally
do identify speakers' illocutionary intents in this way, it may be argued
that the SAS does indeed represent the form of inference hearers make.
However, this line of argument may seem suspect. In particular, what
is the force of saying that hearers have to make such inferences in order
to identify illocutionary intents? Any inference that results in a correct
(or even justifiable) identification will do as far as identifying the illocu-
tionary intent is concerned, but that is not what we want. We want an
inference form that yields justified identifications, one that does so
using information available to hearers in ordinary communication situ-
ations. Our claim is that the SAS represents just such an inference
pattern. To claim further that it represents inferences hearers actually
make requires further claims: (a) that hearers cannot make justified
inferences without using such information, (b) that no other pattern of
inference using just that information leads to justified identifications,
and (c) that hearers' identifications are, at least generally, justified. If
these claims can be established, it follows that hearers generally do
make inferences in the form of the SAS. Otherwise, there would be no
explaining hearers' general success. We consider psychological data
relevant to these claims in chapter 11.
Speech Act Schema 91
Here we should make clear just what we are claiming for the SAS
and what we are not claiming. On the one hand, we are claiming that
the SAS represents both the search procedure and the justification used
by the hearer in identifying the speaker's illocutionary intent. These are
not normally distinguished for a very simple reason: that the hearer
arrives at a plausible candidate for the identity of the intent is, and is
taken to be, good reason to believe it to be the correct one. On the
other hand, although the step-by-step character of the SAS represents
the natural way of organizing, linearly, the abundance of information
available to and relied on by the hearer, we are not claiming that the
process of inference is necessarily sequential in nature. When S's in-
tention isn't transparent and the inference made automatically, H may
have to go through a trial-and-error process of toying with different
interpretations compatible with both S's utterance and with what he
(H) believes about Sand S's likely intentions under the circumstances.
Though operating on the overarching assumption, the CP, that S's
intention can be recognized and is expected to be, H may still need
some imagination to work it out. He may have to go back and reinter-
pret what S said, by changing references or the operative senses of
ambiguous terms. He may decide that S is speaking nonliterally, or he
may conclude that he didn't hear S correctly and try to recall S's exact
words. H may have trouble arriving at a determination of force and
content that are compatible, for one reading of the content may suggest
one way to take the utterance, another another. In practice, of course,
H can ask S to repeat himself or to restate what he said.
5.3. ON THE PSYCHOLOGICAL REALITY OF THE SAS
To suppose that the SAS represents the pattern of inference that hear-
ers make in identifying communicative intentions is to make a rather
strong assumption about the human cognitive abilities involved in
communication. We cannot demonstrate that people normally do make
inferences represented by the SAS, but we will attempt at least to show
that this supposition is empirically plausible. We suggest that inferences
according to the SAS are no more complex than all sorts of other
inferences people commonly make. Moreover, we argue that commu-
nicative inferences must follow the pattern of the SAS if they are to be
rational at all-and we assume they are. Then in section 5.4 we discuss
those aspects of the SAS that rest on an assumption that the role of the
presumptions and of mutual contextual beliefs in communication is no
 ,
The Theory 92
different from the role of mutual beliefs in social phenomena generally.
Our overall goal is to show that the SAS imputes nothing to people,
either cognitively or socially, that is not reasonable to attribute to them
in other contexts than communicative ones. In other words, we suggest
that the cognitive and social aspects of communication are but special
cases of much more general phenomena in cognitive and social psy-
chology.
One thing we are assuming is that people can and do perform com-
plex mental operations very rapidly, generally unaware of the details of
what they are doing. This assumption may be objectionable to those
who reject the idea of mental operations altogether or to those en-
thusiasts of introspection who believe that every mental state is subject
to its possessor's awareness. We know of no general way to appease
dogged behaviorists on the one hand or die-hard Cartesians on the
other and can only challenge them to account for people's ability to
identify communicative intentions without appealing to such complex
cognitive processes as those we are assuming to take place. We believe
that there is nothing unusual about the complexity of the inferences we
attribute to hearers. Equally complex operations seem to be involved in
such processes as perceptual identification, motor coordination, and
problem solving. Of course, it makes little sense to make quantitative
comparisons of this sort, hence there is little content to the claim that
communicative inference is no more complicated than, say, listening to
music or playing tennis. We are in no position to say what any of these
abilities involves; on the other hand, we see no reason to deny that a
detailed scientific account of communicative inference should be of the
same order as that of other human cognitive abilities. In particular, our
cognitive assumption is no more extravagant than that made by many
contemporary linguists and psycholinguists that aspects of the grammar
of a language reflect certain highly complex cognitive abilities involved
in the perception and production of sentences.
Let us turn to the details of the SAS. An inference in the pattern of
the SAS is not deductive but what might be called an inference to a
plausible explanation, namely, of the speaker's utterance. 7 In general,
one good explanation is enough. The inference is abetted by the pre-
sumptions, which license the belief that there is some communicative
intention identifiable from what is uttered and relevant mutual contex-
tual beliefs. The latter provide the basis for determining what that
intention is. The pattern of the SAS plays the dual role of representing
both H's procedure for identifying S's communicative intention and
H's justification for thinking he has identified it. The very fact that an
Speech Act Schema 93
intention has been picked out is usually sufficient to justify the claim
that the speaker has it. However, the SAS gives no detail, at least in the
case of nonliteral and indirect acts of communication, about the spe-
cific strategy the hearer uses to identify a particular communicative
intent. It gives no indication of how certain mutual beliefs are activated
or otherwise picked out as relevant, much less how the correct identi-
fication is made. And yet people do it somehow. So there is even more
to hearers' inferences than what is represented in the SAS.
It would be preposterous to claim that hearers ordinarily8 go through
the SAS explicitly, consciously proceeding from one step to the next.
Were this the rule, we would not be so hard put to specify the sorts of
mutual contextual beliefs that enter into illocutionary inferences, or to
spell out the search strategies that exploit them. But we should not be
too skeptical about the existence and unconscious use of mutual beliefs
and search strategies. Think of the comparably complex and inexplicit
processes (whatever they may be like in detail) involved in such cogni-
tive operations as perceptual identification, motor coordination, and
social behavior. Our empirical thinking in general is rife with generali-
zations and inference principles that we are not conscious of when we
use them, if we are conscious of them at all. It would take us well
beyond present-day cognitive psychology to speculate on the details of
any of this. Instead, we merely suggest that complex, inexplicit opera-
tions are commonplace, that most of our everyday practical thinking
involves much more than meets the introspective eye. Whatever these
processes are, whatever activates them, whatever principles or strat-
egies are involved, they work and work well. Almost all such inference
is implicit. We cannot merely by introspection formulate in any detail
the principles we use, and it is extremely difficult to reflect on the
patterns of one's own past inferences to ascertain what principles or
strategies they embody. Nevertheless, they seem to possess the same
order of differentiation and articulateness that linguistic performance
possesses, and cannot be regarded as being less complex or governed
by rules (principles, strategies) of lesser complexity. That we cannot
readily identify all their ingredients attests to their complexity and
subtlety, not to their nonexistence or blatant simplicity.
5.4. THE SAS AND SOCIAL PSYCHOLOGY
We have suggested that there is nothing inordinate about the cognitive
complexity of the pattern of inference in the SAS. Such inference
involves the activation of mutual contextual beliefs and the application
The Theory 94
of the presumptions, which are themselves mutual beliefs. We claim
that there is nothing extraordinary in the supposition that communica-
tion situations involve such mutual beliefs. Indeed, as argued in the
appendix, social situations in general are governed by mutual beliefs of
various sorts. There it is also proposed that basic sociological concepts,
such as rules, roles, and groups, can be analyzed in terms of certain
mutual beliefs. Here we would like merely to give some indication of
how social situations have certain features describable in terms of
mutual belief, features shared by communicative and noncommunica-
tive situations. Our hope is to show that from the standpoint of social
psychology there is nothing special in kind about the nature of commu-
nicative situations and the fact that they involve mutual beliefs. We will
not pretend to have offered a theory of the social psychological side of
linguistic communication. As with the cognitive psychological side, we
seek merely to make the attribution of inferences in the fashion of the
SAS empirically plausible.
In virtually every interpersonal situation, a person can and does
classify the situation and the persons present as being of certain sorts.
If he is at a ball game, he is aware of that fact and he classifies those
present as players, managers, coaches, umpires, spectators, ushers,
and vendors. There are not only rules of the game proper, but rules
governing. spectator, vendor, and usher behavior. Certain subsituations
are defined as selling, seating, or applauding situations. How a person
is categorized has much to do with what he is to do when. The rules
governing behavior in the various subsituations that arise are by and
large mutually believed by those present, so that almost everyone not
only knows what to do and what not to do when, but has reason to
believe that almost everyone else has like knowledge regarding himself.
What is more, each has reason to attribute this knowledge of everyone
else's knowledge to everyone else. In short, the various rules applying
to types of persons and types of situations are mutually recognized. Of
course, there may also be more specialized rules, not mutually recog-
nized by all concerned, that pertain to such matters as how a player
should wear his socks or how a hot dog vendor should apply mustard.
We suggest that every interpersonal situation, ranging from casual
(or intimate) two-person encounters to elaborate ceremonies and com-
plex institutional activities involve mutually recognized rules that apply
to persons and types of situations. What is especially germane to our
subject is that part and parcel of the existence and sustenance of such
rules are people's expectations, both descriptive and normative, about
Speech Act Schema 95
one another's behavior in various situations that arise in everyday life.
Not only do people expect one another to act in certain mutually rec-
ognized ways, as determined (at least in part) by mutually recognized
rules governing mutually recognized types of persons and types of
situations, they expect others to expect them to act in these ways.
None of this excludes the possibility of either individual differences
in social behavior (generally social rules do not specify all that is to
be done, but impose general requirements that may leave open many
options) or individual violations of the rules. The point is merely
that social situations are invariably, though to different degrees, con-
strained by mutually recognized rules and that people have expec-
tations, recognize expectations, make judgments, and acknowledge
judgments concerning each other's behavior. In particular, they judge
others' behavior, and assume that others judge their behavior, in terms
of these rules, and regard compliance with or violation of them as
intentional. Indeed, not only is this behavior intentional and regarded
as such by others, at least to some extent it is intended to be regarded
as intentional. Not that people always explicitly formulate intentions
about how others are to construe their behavior, though this happens
fairly often. But given their awareness of the situation and the persons
in it and their recognition of the rules that govern it, and given the
mutuality of this awareness and this recognition, people's behavior
vis-a-vis the rules cannot but be expected to be regarded by one an-
other as intentional. Awareness of the situation invokes the rules; rec-
ognition of the rules activates the expectations.
It does not follow that social behavior, though generally intentional
and expected to be recognized as such, is communicative, but it comes
close. Social behavior is not necessarily communicative because peo-
ple need not intend that others make inferences on the basis of their
behavior (much less expect to be recognized as intending it). However,
it is perfectly possible and hardly unusual for social behavior to be
communicative. Rule violations, if committed with recognizable R-
intentions, can communicate the agent's contempt for society, his dis-
respect for those present, or even his acknowledgment that the rules
can be relaxed. Or a person can obey the rules in such a way that he
communicates his contempt for them or for those present. So commu-
nication in the course of obeying or violating social rules is quite possi-
ble. Needless to say, nonlinguistic communication in the course of
social interaction need not be based on obeying or violating social
rules. We do not propose to catalog the variety of such acts of commu-
 ,
The Theory 96
nication, but we do believe they -involve R-intentions and recognition
thereof. Of course, the linguistic and the communicative presumptions
are not relevant. In general, a piece of social behavior is communica-
tive if it has some feature that is R-intended to be recognized and to
serve as a basis for recognizing the agent's intent. Since (generally) no
CP is involved, that feature must call attention to the fact that the
behavior is communicative, that is, that it is R-intended to serve as a
basis for explaining the behavior by identifying the agent's intent. Mter
all, communication, whether linguistic or otherwise, is the upshot of
behavior exhibited with the intention that relevant observers ("ad-
dressees") explain it by attributing a certain (R-) intention to the agent.
Lack of linguistic knowledge and of social knowledge generally
makes it not only impossible to know what to do and not to do when,
but impossible to form communicative intentions or recognize others'
intentions. This is obvious to anyone who has experienced a radically
unfamiliar or alien social situation. Moreover, we are able to identify
people who lack this knowledge. Indeed, part of our knowledge is of
how to categorize those who lack it: as foreigners, morons, or children.
Finally, there are those who seem not just ignorant of but oblivious to
social rules and expectations. It is no wonder that communication to
and from them is especially difficult and frustrating.
Much more could be said about the cognitive aspects of language and
of nonlinguistic social behavior. We have tried merely to indicate that
there is nothing unusual or unrealistic about the attributions to people
made implicitly in the SAS. We have suggested only that nonlinguistic
communication is an everyday affair and that it involves complex be-
liefs, intentions, and inferences. It may be somewhat less complex than
linguistic communication, but only because of the complexity of lan-
guage itself and of the resultant richness and variety of linguistic acts.
5.5. COLLATERAL ACTS AND SOCIAL MOVES
We have so far given the SAS center stage in the discussion of talk-
exchanges, mainly because linguistic communication is one of the pri-
mary purposes of such exchanges and, if we are right, the SAS is
central to linguistic communication. But not all linguistic acts are per-
formed via the SAS. In this section we survey noncommunicative lin-
guistic acts and classify the intentions with which they are performed.
Moreover, inasmuch as talk-exchanges are social as well as linguistic
events, we think it useful to mention some of the ways a talk-exchange
may be advanced, as well as the conditions necessary to sustain it.
Speech Act Schema 97
5.5.1. Collateral Acts
An assortment of conversational acts can be performed in conjunction
with or in lieu of illocutionary acts. Let us call these collateral acts.
Like the perlocutionary intentions discussed in chapter 4, the inten-
tions behind these acts mayor may not be intended to be recognized. In
some cases the intention can be fulfilled only if it is not recognized; in
other cases the intention is an R-intention and thus cannot be fulfilled
without being recognized. We might call the first sort covert acts and
the second overt. And some collateral acts involve intentions whose
fulfillment is independent of their recognition; these are neither covert
nor overt in the senses just defined.
Collateral acts that exploit the communicative presumption in one
way or another are necessarily overt, since to succeed the hearer must
recognize this exploitation. Kidding, storytelling, joking, punning,
mimicking, and reciting are essentially overt collateral acts at least
when the addressee is among those who are intended to appreciate the
act. Of course, you can kid or mimic someone without intending him to
recognize it.
To kid someone is, roughly, to R-intend to say something without
meaning it. Suppose, in a romantic situation, S says to H, who is
affectionately tickling her toes but with nearly excruciating results, "If
you don't stop that, I'll tell my mother to stay home." H's mother-in-
law has been invited for dinner, you see, but considering the mutual
contextual belief that H is not excited about this prospect, S intends it
to be obvious to H that she does not mean what she is saying, that she is
not making a threat. Besides, it is mutually believed that S has a low
tickling tolerance, so there is no reason for H to think that S intends to
be informing him of anything. Rather, S is simply kidding-she does
not mean what she says and she R-intends H to recognize that. When
she really wants H to stop tickling her, she will withdraw her feet.
If a speaker states that he is about to tell a story or a joke, he
has thereby suspended the communicative presumption explicitly, R-
intending his audience to recognize the sequel for what he is stating it to
be. But suppose there is no explicit indication that the CP is being
suspended, so that the audience has to recognize the utterance as a
story or ajoke. S R-intends H to do this, expectingH to realize that S's
utterance is not a genuine illocutionary act. That is, what S is saying
cannot be taken, either literally or figuratively, directly or indirectly, as
expressing any attitude on the part of S, including the belief that his
utterance is an expression of an attitude (it is not necessary to the
The Theory 98
successful telling of a story or a joke that one be taken as believing or
intending it to be edifying or amusing). S R-intends H to take his
utterance to be simply a presentation of something fictive; presumably
this intention is fulfilled because of the content of S's utterance. S is
exploiting the CP by violating it to the extent that he is not performing
an identifiable illocutionary act but not to the extent that what he is
doing is not identifiable. The content makes it identifiable as a story or
joke. Of course, the story or the joke may itself require the audience to
make certain inferences, but these reflect the intentions of the origi-
nator of the lines. In general, it is irrelevant whether the originator
happens to be the speaker.
Puns are a special case of jokes for two reasons. For one thing, they
exploit linguistic relationships, be they phonetic, semantic, or even
syntactic. Moreover, unlike jokes, puns generally occur within regular
discourse. Suppose after finishing his sausage S says to his luncheon
companion at the Dusseldorf Deli "That was the worst thing I could
have done for my liver." He may not have meant what he said, but only
that the liverwurst tasted terrible enough to make him sick, at least
figuratively speaking. As for the pun, his R-intention is that his com-
panion recognize the phonetic connection between "worst . . . liver"
and "liverwurst." Not that a pun must be intended to be recognized. In
any case, especially with a pun like this one, there is always the out,
usually disingenuous, "No pun intended." A known punster, however,
may rely on his reputation to produce plays on words that would be
.undetectable but for the presumption that he makes them at all costs.
Indeed, a punster can exploit his own reputation (and that of puns
generally) by producing puns so trite, trivial, or otherwise outrageous
that his R-intention can only be (one hopes) to produce a pun that is
amusing precisely because it is bad. This is verbal high camp.
Then there is the overt collateral act of mimicking. When a person
imitates someone else's manner of speaking, assuming he has not an-
nounced he is going to mimic, he R-intends the audience to recognize
his intent by means of the similarity between how he says whatever he
says and how the person in question characteristically speaks. Even
when the content of the utterance is relevant, as when Nixon imitators
would begin by saying, "Let me make one thing perfectly clear," the
speaker need not mean anything or intend to be performing any illocu-
tionary act, though he does have an R-intention; in this case the inten-
tion includes that H recognize that what S is saying is what the person
imitated typically says.
Speech Act Schema 99
Finally, for the overt act of reciting (a poem), as well as for acts like
quoting and playacting or rehearsing, the context is normally such,
either in virtue of some prefatory remark or of the situation itself, that
no one takes the CP to be operative. In these contexts it is mutually
believed by all concerned that the lines of a poem or play are being
uttered. Still, it is a matter of R-intention, however transparent, that
this is what is going on.
Certain collateral acts are not necessarily either covert or overt. It is
indifferent to their success whether the intention with which they are
performed is recognized. This is true of many perlocutionary acts as
well, but the collateral acts to be examined here are not perlocutionary
in the strict sense of the term.
In circumlocution the speaker compromises the presumption of man-
ner in order to avoid the offense to the hearer or the embarrassment to
himself that explicit language would engender. Instead of saying "It's
about time you got here," one might say to a tardy guest "We were
beginning to worry about you." Both utterances acknowledge the
guest's late arrival, but only the second does so tactfully.
Changing the subject is a common collateral act with a range of
possible purposes. One may change the subject to avoid revealing a
secret, to keep from committing oneself on something, to avoid exces-
sive dwelling on a subject painful to oneself or to the hearer, to confuse
the hearer, to test the hearer's interest or persistence, or simply to liven
up the conversation. The speaker may have an R-intention, namely, to
implicate by violating the presumption of relevance his reason for
changing the subject, but his intention need not be recognized to be
fulfilled. Whether the hearer cooperates depends on more than his rec-
ognition of the speaker's intention to change the subject.
And there is the phenomenon of small talk, conversation whose pri-
mary purpose is to ease the awkwardness of silence. Here the un-
expressed intention is to fill the air and pass the time with a minimum of
discomfort. This does not imply that the speakers' illocutionary inten-
tions in saying whatever they say are insincere, though the exchange
may lack the importance they are cooperatively pretending it has- you
do not have to care to be sincere. The main thing is that everyone
recognizes, perhaps recognizing that others recognize, the vacuum-
abhorring nature of the conversation, but no one is willing to be explicit
about this recognition for fear of embarrassing himself or insulting the
others.
The social management of embarrassment is a topic for social psy-
 ,
The Theory 100
chology; we will only point out a general feature of it. There are many
things that when mentioned are embarrassing, to either the speaker or
the hearer, even though they are obvious to both: an unsightly facial
blemish, a nervous tic, a foul odor, a faux pas. By keeping touchy
topics verbally concealed, people cooperate to protect their social
selves from one another. Any utterance that focuses attention on
something else is collaterally the act of covering up a source of embar-
rassment. The intention to do so may be intended not to be recognized,
but it need not be so intended.
The examples just mentioned are special cases of a pervasive social
phenomenon, the verbal taboo, a rule against mentioning a certain
thing. Something need not itself be taboo for mentioning it to be. For
example, there seems to be a taboo against mentioning certain features
of social relationships. In a doctor-patient relationship the ground rules
may permit the doctor to touch, and prohibit the patient from prevent-
ing him from touching, the patient's private parts. Despite the lack of
personal familiarity between the two, this physical contact is regarded
as perfectly proper-most patients are not overly embarrassed by it,
and most doctors do not feel that they are intruding. Nevertheless,
there seems to be a social dictate against mentioning the nature of this
situation; to do so would be embarrassing. Similarly, when people of
unequal social or professional position have occasion to interact, even
though the inequality is mutually recognized as the superior exercises
his rights and enjoys his privileges over his inferior, it would be socially
awkward for either to mention the inequality of the situation. Not only
would that be an act of disrespect, it would threaten, or be perceived to
threaten, the social structure in which this relationship obtains. As in
the cases of cooperative coverup, in an utterance issued with the col-
lateral intention of avoiding mention of and diverting attention from a
verbal taboo, the intention need not be covert to be fulfilled. Verbal
taboo does not prevent the unmentionable item from being an object of
awareness, but it protects all concerned from having any responsibility
for making the awareness explicit, much less mutual.
Motives of politeness, respect, and self-protection figure in a variety
of collateral acts. Social grace may require speaking in a certain style,
evidenced by using certain forms of words and certain modes of pro-
nunciation and inflection. Forms of address can be carefully chosen,
and the do's and don'ts of social propriety can be carefully observed.
To some extent the motive of social propriety may be evident from a
person's illocutionary intentions, but largely one's intention to be and
Speech Act Schema 101
to appear proper is collateral to the performance of particular illocu-
tionary acts. This intention, whether recognized or not, is fulfilled more
in the how than in the what of one's utterances and other acts of social
exchange. The intention to be or appear proper (or improper, offensive,
cool, hip, sophisticated, powerful, important) cannot be an R-intention,
although one may wish to communicate that one has knowledge of how
to be proper (or whatever) and utter something with an R-intention to
this effect. But the intention to be or to appear proper cannot be overt
and, if it is to be fulfilled, may even have to be covert. For in some
circles part of what being proper involves is not having to make an
effort, hence not having to intend, to be proper.
Covert collateral acts are performed with intentions that are intended
not to be recognized. Generally, these are acts of manipulation, in-
cluding such devious acts as innuendo, deliberate ambiguity, and
"sneaky presupposition." Whereas indirect illocutionary acts are not
explicit and yet are performed with R-intentions, covert collateral acts
succeed (the intention with which they are performed is fulfilled) only if
their intent is not recognized, or at least not recognized as intended to
be recognized. The idea is to get someone to think you think something
and thereby to get him to think it without recognizing that that's what
you want him to do. Whereas an indirect act is performed with an
intention that can be. reasonably expected to be recognized (on the
basis of the utterance and the context), so that the speaker cannot, if
challenged, plausibly deny that he intended the hearer to infer his in-
tention, the key to innuendo is deniability. One can make a veiled
suggestion that someone is a foreign agent by saying "Sparsky didn't
look the least chagrined when Krasny told him that Azevedo's Portu-
guese cover had been blown." If this sentence is uttered in the course
of a matter-of-fact account of an observed conversation between
Sparsky and Krasny, both of them on the same side as the speaker and
his audience, the inference that Sparsky is a traitor is not going to be
drawn on the basis of recognizing an R-intention. The speaker knew
this in choosing his words and bears no responsibility for the inference
he covertly intends to be drawn.
Deniability is preserved also in what Ann Weiser (1974) calls "delib-
erate ambiguity." If Boy says to Girl, "Are you doing anything to-
night?" he protects himself from the embarrassment of rejection by
intending his utterance to be taken merely as a question if her answer is
"Yes" and as an invitation if it is "No." Assuming the hearer does not
recognize the speaker's intention to be ambiguous, the speaker has
The Theory 102
direct control over how the utterance will be taken, given the response.
If the hearer does recognize his intention, she can determine how the
utterance has to be taken, by saying such things as "Yes, but I'm free
tomorrow night" or "No, I'd rather stay home and read tonight." In a
later article Weiser (1975) calls the move of taking the ambiguity one
way rather than the other "selection by reply."9
In deliberate ambiguity clearly the speaker does not have an R-
intention that his utterance be taken in one of two ways. Rather, since,
his deliberate ambiguity is covert, he has a simple intention that the
hearer take him to have one of two possible R-intentions, which one to
be determined by the subsequent course of conversation. Weiser (1974)
emphasizes illocutionary ambiguity in her discussion of devices for
managing socially tricky situations, but semantic ambiguity can work
equally well, as in "I'd like to see more of you" uttered under suitable
circumstances. Of course, deliberate ambiguity need not be covert, but
then its motive is not self-protection but titillation, as "I'd like to see
more of you" would be if uttered with rising intonation and eyebrows in
the manner of Groucho Marx. Double entendre is semantic ambiguity
that is intended to be recognized, as when S, admiring H's gilded
coathangers, says "I'd love to hang up your clothes for you," which
can be taken as both an offer and an invitation.
Another kind of covert collateral act is what Hutchinson (1971)
speaks of as "sneaking new information into presuppositions." For
example, one might say "Fortunately, the CIA is no longer involved in
political assassinations" in order to avoid asserting explicitly that the
CIA used to be so involved and thereby to make it more difficult for the
hearer to challenge this proposition. To cover for the possibility of
error or refutation, instead of making a bald assertion one might ask a
question like "Did you know that there are abstract performatives?"
intending to agree with the hearer whether he says "Of course I know
that!" or "Whad'ya mean-are there unicorns?" In the latter case the
speaker is prepared to cancel the implication that he believes in ab-
stract performatives by saying "I only wanted to see whether you've
fallen for that line." Generally speaking, whatever the speaker's reason
for slipping something into a presupposition, whether it is to avoid
committing himself, to protect himself against being challenged, or to
be verbally economical, his intention is most effective if it can be co-
vert. But even if he expects the hearer to recognize his intention, still
he has succeeded in putting the onus on the hearer to question the
presupposition or admit ignorance of it.
Speech Act Schema 103
No doubt there are other kinds of collateral acts, be they overt,
covert, or neither. Our object has been merely to give some common
examples of each of the three types.
5.5.2. Conversational and Social Moves
A conversational situation is a social situation. Certain sorts of speech
acts are essentially concerned with the course or direction of the social
situation. Some of the acts listed in our taxonomy ofillocutionary acts,
such as responsives and acknowledgments, are of this sort, but situa-
tionally oriented acts are not limited to illocutionary acts. There are
two general kinds of such acts, those that presuppose certain specific
features of the sociaVconversational situation and those that directly
affect these features. Acts of the first kind help sustain the sociaVcon-
versational situation by keeping it within mutually expected bounds,
either by being appropriate to a given stage of the exchange or by
carrying the exchange to the next stage that is mutually expected. Acts
of the second kind do not sustain but restructure the situation. Their
purpose is to change the terms or the course of the exchange, and their
success requires cooperation or at least lack of opposition from the
hearer.
Both kinds of acts invoke certain mutual beliefs about the nature of
the situation. If the mutual belief invoked is intended by the speaker to
be invoked and intended to be recognized as so intended, the situation-
ally oriented act is overt- the speaker cannot but expect the hearer to
refer to this mutual belief in identifying his illocutionary intent; thus the
illocutionary intent is identified in terms of the invoked mutual belief,
and so the act expressing one's intention to sustain or to alter the
situation is an illocutionary act. However, since more than uptake is
required for this intention to be fulfilled, the act of actually sustaining
or of altering the situation is perlocutionary. Perhaps a few examples
will clarify these remarks.
Grice (1975) has suggested that "talk-exchanges" are governed by a
cooperative principle. under which fall various conversational pre-
sumptions (he calls them maxims). They must be mutually recognized if
they are to apply to a given conversationaVsocial situation, and can
thus be viewed as rules (see section 7.1). They are fairly general in their
application, although, as Grice notes, his maxims have to be general-
ized to cover exchanges whose purpose is not merely conveying infor-
mation. Particular social situations are governed by whatever rules the
 The Theory 104
participants mutually believe to govern them-not that there is always
agreement on what they are. To the extent that there is less than mutual
agreement, the situation is ill defined and one or both (all) of the par-
ticipants may be unsure of what to expect from the other or unclear
about what the other expects of him. What social psychologists some-
times call "role negotiation" may occur here if the participants take
measures to clarify the situation and their places in it.
The rules that govern a situation, assuming they are mutually recog-
nized, may be determined by the generic character of the situation as
belonging to a certain institutional procedure. In this case the primary
determinant of what the situation is (the rules of the institution) are
mutually recognized among the members at large. The personal re-
lationship of the participants may thus be incidental, and their relative
institutional positions may be the decisive factor. Even when an en-
counter is of a more personal nature, with considerable intimate
knowledge between the participants, it cannot be assumed that the
rules governing the encounter are idiosyncratic to that relationship.
The participants may still categorize each other in socially recognized
terms and mutually define the situation as being essentially, say, a
student/teacher situation, governed by socially recognized rules. How-
ever, insofar as the participants have developed a relationship of rela-
tively long standing, it is likely that they have developed a certain
mutual personal understanding, including an understanding about what
is required or permissible in various recurrent situations that they have
learned to classify in certain ways: as a neighborly visit, a coffee break,
or a lovemaking scene.
However a situation is defined and whatever the source of the rules
that govern it, by and large there is at any given moment a mutual
awareness of how the situation is proceeding and what sorts of acts are
appropriate to its furtherance. For example, a neighborly visit has to be
initiated somehow-the visiting neighbor does not (usually) just walk
in the door, sit down, and start talking. Each party is likely to have a
certain view of why visits are made (say for idle conversation), how
they should be initiated, what is a good reason for refusing a visit, and
how frequent, long, and involved visits should be. So the visitor, after
the exchange of greetings, might (especially if he is not invited in im-
mediately) ask the host if he's busy at the moment. After he enters, he
may presume to sit down, perhaps wherever he likes, or it may be
mutually expected that the host ask him to sit down. Clearly, what is
mutually understood affects how a given act will be taken. If it is usual
Speech Act Schema 105
for the host to ask the visitor to sit down, sitting without being asked
might be regarded as presumptuous or offensive. Indeed, it might be
regarded as being intended to be regarded as such. On the other hand,
the visitor might so presume in order to make the relationship less
formal. Tactfully, he might say as he sits "I hope you don't mind,"
inviting his host to say "Of course not" and thereby making future
permissions to be seated unnecessary. If this becomes the established
state of affairs, then for the host on some future occasion to ask the
visitor to sit down suggests a preference for less intimacy and more
formality. Analogous patterns emerge regarding such aspects of a visit
as taking off one's coat, serving (or helping oneself to) refreshments,
turning on the TV, and so on. Finally, there is the matter of terminating
the visit. Neither party is likely to say he's bored with the conversation
or would rather just be alone, but either can mention a real or plausible
commitment or obligation that justifies taking leave. Then things can
end as they began, on mutually agreeable terms.
Conversational situations are never just conversational. They are
governed by social rules as well as conversational rules. Insofar as
these are mutually recognized-whether institutionally imposed, de-
termined by the sorts of persons involved, or personally imposed and
reflective of the individuals involved-they provide guidelines within
which acts (linguistic and otherwise) are performed and perceived.
Compliance with the rules, unless it is unexpected,provokes no special
attention and invites no special inference. However, violation does
more than raise eyebrows. It calls attention to itself and invites infer-
ence as to why it occurs. Perhaps the point of it is civil disobedience:
violating a rule in order to call attention to the rule and change it. More
common, though, are violations designed not to change rules but to
change the nature of a situation. The person objects not to certain rules
governing certain situations but to a particular situation's being of the
sort to which a certain set of rules applies. He seeks to change the
situation to one with different rules.
What counts as relevant to a conversation depends on its mutually
accepted purpose or direction. But this can change-or be changed-
during the course of a conversation. There are rules that determine ac-
ceptable ways in which the course of social situations can be changed.
A person might want to change the subject abruptly in midstream;
perhaps he is bored with it, perhaps it is getting too touchy for him.
Although it may be socially proper to change the subject without giving
reasons for so doing, sometimes reasons are required, as when the
The Theory 106
subject is of obvious interest and importance to the other person. On
the other hand, it is socially improper to say that one is bored and at
least awkward to say the subject is getting too touchy-saying so
would only make it touchier. So one must find an acceptable way of
changing the subject without conveying one's real reason. Fortunately,
it is possible to change the subject without one's intent being recog-
nized and without its being recognized as one's intent. A possibly ploy
is simply to change it, perhaps with an innocent-sounding "Oh, by the
way, did you know that . . . ?" One relies on the likelihood that even if
the other recognizes that the subject is being changed intentionally and
even if he is thereby piqued at this, he won't accuse one of willfully
changing it. Such accusations are themselves not socially acceptable,
especially if they aren't backed up. And unless one repeatedly changes
the same subject, one can always plead innocent, denying any vicious
intentions. Another move, not without its irony, is to use a disclaimer
like, "I don't mean to change the subject, but did you know that ... ?"
Here the person changes the subject explicitly and intentionally, at the
same time paying obeisance to the right, which he is in the process of
violating, of the other that he stick to the subject. Moreover, he puts
the onus on the other to call him to task for what he is doing.
Most of the time changing the subject is perfectly acceptable. Most
conversations are not all that regimented and subjects are dropped right
and left in the normal course of events. Nevertheless, some discretion
is required. Questions should be answered (or be allowed to be an-
swered), points should be allowed to be completed, and civil attention
should be paid. But all this is not so much a matter of relevance as
politeness.
The countermove to changing the subject is to change it right back,
perhaps after giving the changer the benefit of the doubt (maybe his
mind was just wandering). This can be done bluntly and without any
explanation or by using such words as "Getting back to what we were
talking about," "Before I forget what we were discussing ... ," or
"Anyway, as I was saying ... " This can be done tactfully, of course,
but if persistence is required, it may not be appreciated. However, the
changer of the subject bears the onus of having to change it again or of
finding a suitable way to accuse the other of obnoxiously pursuing a
tiresome topic.
Other Gricean maxims (presumptions) can be overridden by social
rules. For example, the presumption of quality might require telling the
truth but propriety might demand a white lie. The presumption of
Speech Act Schema 107
quantity might require considerable detail (perhaps required also by the
presumption of quality), and yet going into those details might be
boorish or just plain boring to participants in the conversation. The
presumption of manner might be overridden by various social consid-
erations; for example, it might be necessary to be ambiguous (Weiser
1974) in order to provide the hearer with options that, if not taken up,
one has not openly committed oneself to-as in "Are you doing any-
thing tonight?" Again, one might have to violate the presumption of
orderliness if, for example, the situation is so informal that any attempt
at order would seem inappropriate. In general conversational situations
are not merely conversational, and conversational presumptions can be
overridden not just by each other but by social rules.
Conversational scenarios, even when not multilayered and emotion-
ally charged, are governed not simply by (strictly) conversational pre-
sumptions. Conversations are social encounters, encounters that have
to be initiated, furthered, and terminated; sometimes they can be pro-
longed, sometimes they must be abbreviated. Although many linguistic
devices are suitable for affecting the course of a conversation, the rules
governing their use are not limited to purely conversational presump-
tions. There may be rules, depending on the nature of the situation,
governing who can start a conversation-in some situations the rule
may be that one speak only when spoken to. The same point applies to
terminating conversations. Authority relationships may determine not
only who can start and end a conversation but who directs it and how.
Even in casual exchanges between equals, there are principles of pro-
priety and a catalog of offenses that are to be avoided in initiating,
directing, or terminating a conversation. These are not exclusively
conversational in nature and indeed may conflict with strictly conver-
sational presumptions.

Chapter Six Conventional
Illocutionary Acts
The speech act schema and the pattern of analysis for communicative
illocutionary acts do not apply to conventional illocutionary acts. A
different story is needed for them. Seeing why they do not fit the SAS
will help one appreciate the explanatory value of the SAS for com-
municative acts. We will see just how conventions take the place of
R-intentions in determining that utterances count as performances of
particular kinds of acts. Whereas a communicative intention is fulfilled
by means of recognition of that intention, a conventional intention is
fulfilled by means of satisfying a convention.
Conventional illocutionary acts include such diverse acts as voting,
resigning, arresting, acquitting, marrying, christening, dedicating, and
abolishing. Despite their diversity, they fall into two general categories,
effectives and verdictives, both of which affect institutional states of
affairs.
6.1. CONVENTIONS
The idea of conventionality is quite broad; our conception of conven-
tion captures, and is meant to capture, but one part of it. For us con-
ventions are counts-as rules and nothing else. When we talk about
conventional illocutionary acts, we will be referring to what makes
utterances count as acts of certain sorts. Conventional illocutionary
acts need not be ritualistic, artificial, formal, fashionable, orthodox, or
anything else that might fall under the broad notion of conventionality.
We hope that once mentioned, these ideas won't be confused with what
we mean by convention. Confusion is possible not only because the
term conventional applies to so much, but also because these different
aspects of conventionality often go together. Our notion of convention
Conventional Illocutionary Acts 109
. not to be confused with any of the conceptions of rules that have
ffl .
been bandied about in the philosophy of language (such as semantic
rules and constitutive rules) nor with social rules, practices, or non~s.
As we understand them, conventions are not, or are not necessanly,
what people expect one another to do in certain situations. Rather, they
are actions which, if done in certain situations, count as doing some-
thing else. In other words, a convention is a mutually recognized means
for doing something, counting as such only because mutually recog-
nized, perhaps by having been agreed upon. Conventions of this s~rt
occur, for example, in the game of bridge: the Blackwood convention
and the Stayman convention, for instance. Bridge conventions are
conventional means, within the rules of bidding, for requesting or con-
veying information. Similarly money is a conventional means of ex-
change. And it is a commonplace, however unexplicated, that language
is a system of conventional means for communicating. The explication
that we offer spells out the ingredients essential to the existence of a
convention, namely, by virtue of mutual belief (MB) in a community or
group (G) an act of a certain sort (A) counts as doing such and such (D)
in a certain sort of recurrent situation or context (C):
Convention: A (in C) is a convention for D-ing in G if and only if:
i. it is MB-ed in G that whenever a member of G does A in C, he is
D-ing, and
ii. A in C counts as D-ing only because it is MB-ed in G to count as
such.
The second condition is necessary to exclude cases of mutually recog-
nized habitual action, like the act of trying to get warm by putting on a
coat. Thus, to A conventionally counts as D-ing not just for happening
to be, but because it is mutually regarded as D-ing. Specification of the
type of situation is essential: shaking hands is a way of greeting in one
situation, a way of bidding farewell in another, a way of congratulating
in a third, and a way of sealing an agreement in still another. In general,
only in certain circumstances do certain kinds of acts count as baptizing
someone, stealing a base, saluting an officer, calling a meeting to order
(using a gavel), or signaling a touchback.
Before applying our definition (which is intended as an explication of
but one sense of convention) to illocutionary acts, we wish to make two
general observations. First, there can be rules requiring the perform-
ance of a conventional action such as saluting, paying taxes, or toasting
someone. Where such rules exist, they must be clearly distinguished
The Theory 110
from the conventions defining the actions they require. 1 Second, the
notion of convention, defined for our purposes over types of acts, can
be extended to types of persons, situations, and objects: being a spouse
or a judge requires having some mutually recognized feature that may
include having performed some conventional act (like vowing) or hav-
ing been the subject of such an act (like being appointed). Analogously,
a situation's being a trial, a debate, or a wedding is a matter of conven-
tion, and the same goes for a thing's being a uniform, a stoplight, or a
crucifix.
Communicative illocutionary acts succeed by means of recognition
of intention, whereas conventional ones succeed by satisfying a con-
vention. Utterances are the A's in the definition of convention, and
what they count as (the D's) depends on the conventions they fall
under. In some cases the convention requires a specific form of words,
in other cases not. For example, a voice vote must be cast with an
"aye" or a "nay," and a verdict must be brought in with the words
"guilty" or "not guilty." An oath of office or a marriage ceremony
requires specified sentences to be pronounced by the participants. We
may call conventional illocutionary acts of this sort locution-specific. In
other cases, what must be uttered is identified by the meaning of what is
said, not by the specific form of words. It does not matter whether a
boss fires an employee by saying "You're fired" or "You are hereby
relieved of your duties and their attendant remunerations."
For an utterance to be a conventional illocutionary act, not only
must it be the utterance of what the convention requires (the specified
words, or words with the specified meanings), it must be issued by the
right person under the right circumstances. Not just any utterance of
"guilty" counts as finding a person (defendant) guilty. It must be said
by the judge (or the foreman of the jury) at the appropriate stage of the
judicial proceedings. Thus, where it says in our definition of convention
"whenever a member of G does A in C," it should be understood that
the specification of C may include a requirement on who the agent is.
This could also be included under the specification of A, but for con-
venience in discussing conventional illocutionary acts we will let values
of A be of the form "says that *( ...p ... )" or "utters e," in the case of
locution-specific acts.
6.2. EFFECTlVES AND VERDICTlVES
Conventional illocutionary acts come in two categories, effectives and
verdictives. 2 Effectives effect changes in institutional states of affairs;
conventional Illocutionary Acts 111
they are necessarily conventional inasmuch as they achieve their ef-
fects only because mutually believed to do so. Only thus is a student
graduated, a bill vetoed, or a site consecrated. Verdictives are judg-
ments that by convention have official, binding import in the context of
the institution in which they occur. Thus, to call a runner out, to find a
defendant guilty, or to assess a piece of property is not just to make a
judgment; given the position and attendant authority of an umpire, a
judge, or a tax assessor, it is also to make it the case, if only so far as
the relevant institution is concerned, that what is judged to be so is so in
fact.
Generally speaking, conventional illocutionary acts, whether effec-
tive or verdictive, are endemic to particular institutions. In most in-
stances they affect the institutional status of persons or things. In other
cases they create institutional rights and obligations. In still others they
further or are otherwise part of some institutional practice, process, or
procedure. We will not attempt an exhaustive enumeration of such
acts. That may well be a futile effort, considering the variety of institu-
tions there are. Things would get out of hand if we were to include
conventional acts performed in writing-for starters, think of the num-
ber of forms there are in bureaucracies, forms for reporting, billing,
requisitioning, applying, notifying, and on and on. Our sampler will be
restricted to fairly familiar acts designated by simple verbs. Generally
they are effectives; verdictives will be marked with a "V." It should be
kept in mind that many of these acts can be, and in some institutions
must be, performed nonverbally. Sometimes words accompany the
nonverbal action, as with christening a ship.
A great many conventional illocutionary acts affect the institutional
position or social status of a person. He may be admitted into an in-
stitution or a position in it by being hired, appointed, nominated,
elected, promoted, naturalized, or ordained; he may be removed by
being fired, suspended, demoted, expelled, or banned. He himself can
enlist, apply,join, or accept membership, and resign, abdicate, retire,
or take a leave of absence. Formal or ceremonial acts can make such a
change of position official, as in installing, inaugurating, graduating,
confirming, and administering or taking an oath of office. With respect
to the criminal law, one is arrested, indicted, convicted or acquitted
(V), sentenced, pardoned, paroled, or reprieved. One can be certified
(V) as to competence in a field, social pedigree, or sanity. Finally, one
can be disqualified (V), blacklisted, censured, or, for that matter,
cleared (V).
The institutional status of objects is affected by a variety of different
 . ,
f
The Theory 112
acts. As to ownership, things can be bought, sold, borrowed, lent,
traded, donated, conferred, awarded, bequeathed, bid for, put up for
sale, accepted, or rejected. Property can be appropriated, expropri-
ated, repossessed, surrendered, or deeded. Items can be consigned,
supplied, received, designated (for some purpose), or relegated (say, to
the junk heap). Uniforms, emblems, flowers, and colors can be adopted
as official symbols. Special institutional status is conferred when some-
thing is dedicated, consecrated, or enshrined. For that matter, people
can be canonized, memorialized, inducted (into a hall of fame), or
knighted. These honorific acts are not verdictive, although they are
predicated on highly positive evaluations. Essentially verdictive are
acts of appraising, assessing, grading, ranking, and estimating the value
of something.
Then there are institutional permissions, prohibitions, and require-
ments, which create or remove rights or obligations, perhaps by im-
posing or abolishing rules. Kinds of acts can be sanctioned, licensed,
exempted, prohibited, legalized, or banned. Particular acts can be
permitted, authorized, enjoined, assigned, delegated, commissioned,
commanded, countermanded, or excused. Rules and permissions them-
selves can be adopted, decreed, repealed, revoked, abolished, or re-
scinded. And states of affairs, such as emergencies and holidays, in
which certain rules or permissions apply, can be declared, proclaimed,
or declared over.
It would be a virtually endless task to enumerate the types of
conventional acts that are part of or further the procedures of such di-
verse institutions as hospitals, schools, government agencies, factories,
churches, and clubs. Take just the familiar and standardized case of
parliamentary situations. Motions can be introduced, seconded, voted
on, or vetoed. Points can be raised, objected to, ruled on (V), adjudi-
cated (V), or overruled (V). People can seek recognition and be recog-
nized, and meetings can be called to order, recessed, or adjourned.
A special case of conventional act involves symbolizing or cate-
gorizing, linguistically or otherwise. There are acts of naming, ab-
breviating, coding, classifying (information), labeling (products), and
such ritualized acts as baptizing, christening, and dubbing.
Finally, institutions themselves can be subject to conventional ac-
tion. They can be legalized, banned, exempted (as from a law), admit-
ted (to an association). Indeed, they must be founded or organized in
the first place, and can be dissolved or disbanded. Acts of founding an
institution require the formation of mutual beliefs among the founding
Conventional Illocutionary Acts 113
members; when an institution is disbanded, there is a general mutual
belief among the remaining members that the rules and expectations
that make up the mutual beliefs within the institution will dissolve.
6.3. CONVENTIONAL ILLOCUTIONARY ACTS AND
INSTITUTIONAL FACTS
Effectives produce or alter institutional states of affairs. Since they are
mere utterances, this essential feature may make them seem ontologi-
cally problematic and invites the question, How are effectives possi-
ble? Verdictives are determinations of fact, natural or institutional,
which have official consequence. Subsequent institutional activity pro-
ceeds as if what has been officially determined to be so is in fact the
case. As an aid to understanding both sorts of acts, we will formulate a
distinction between natural and institutional facts and thereby explain
why effectives and verdictives must be conventional.
6.3.1. Effectives
Effectives are utterances that, when issued by the right person under
the right circumstances, make it the case that such and such. This is a
matter not of causality but of mutual belief. An utterance counts as an
act of a certain sort in virtue of being mutually believed to be an act of
that sort. Only thus does it count as an act of resigning, bidding, veto-
ing, seconding, exempting, or bequeathing, as the case may be. How-
ever, as is evident from a case-by-case analysis of what these acts
essentially involve, to mark such an act with an effective verb is just
shorthand for describing it as fact-producing. Resigning just is remov-
ing oneself from a position, vetoing just is nullifying a piece of legisla-
tion, exempting just is making an exception to a rule, and so on. Of
course, not every kind of institutional fact-changing act is designated
by a simple verb, but every effective verb can be paraphrased in in-
stitutional fact-changing terms. Since to perform an effective just is to
issue an utterance (in a situation) which is mutually believed to be such
that a certain institutional state of affairs is thereby produced, if there is
anything mysterious about effectives, it would seem that they are no
more mysterious than the institutional states of affairs they produce.
What are institutional facts? As argued in the appendix, they are
intersubjective in character, constituted by mutual belief. We might
define them quasi-inductively: an institutional fact is (1) anything that is
The Theory 114
the case in virtue of being mutually believed in some collectivity or
(2) anything that follows from one or more institutional facts.3 For
example, the existence of some obscure law is a matter of institutional
fact whether or not anyone has heard of it, much less whether everyone
mutually believes in its existence. However, it is mutually believed that
there is a legislature which enacts laws (perhaps as provided for by the
mutually recognized constitution) and that its members are determined
by election. So the obscure law exists in virtue of meeting certain
mutually recognized conditions, namely, being passed by the duly con-
stituted legislature (perhaps also being signed by the executive). 4
In general, given the basic institutional fact of the existence of the
institution itself, institutional facts concern positions, rights, respon-
sibilities, and obligations of persons, as well as the actions that people
take and the conventional effects of these actions. Within a particular
institution there may be a structure of diverse official positions whose
occupancy is determined by certain recognized procedures . There will
be rights and responsibilities associated with each position, as well as
rules imposing obligations on everyone. The rules may specify rewards
for exemplary performance or punishments for deficient performance.
Included or implied by the rules specifying rights and responsibilities of
particular positions are power r{(lations between positions, rules that
specify who can (or must) do what to whom and who must tolerate
what from whom (see the appendix). If the institution is organized
toward certain ends, presumably it will have a set of policies and pro-
cedures for furthering those ends. These procedures may be extremely
complex, requiring a huge number of coordinated actions by many
people in diverse positions. Furthermore, there are institutional facts
about these procedures, such as how far along a certain procedure has
progressed (in the legislative process, on the production line) and the
current state of some proposal or production. s
To mention the main matters of institutional fact is but to scratch the
surface of social metaphysics. We have answered our question about
the possibility of effectives only by subsuming it under the larger ques-
tion of the possibility of institutional facts. No one will be content with
the answer that institutional facts are possible because of mutual belief,
because there is still plenty of room for puzzlement about how facts can
exist in virtue of being mutually believed. At any rate (to pass the
metaphysical buck) effectives exploit conventions (one sort of institu-
tional fact) that count utterances, because of what is uttered or said, as
producers of institutional facts. Descriptions of conventions make ref-
Conventional Illocutionary Acts 115
erence to persons and situations, and these involve specifying the in-
stitutional positions of the speaker and others and the relations between
them, as well as institutional facts (such as stages of a process or
procedure) that define the situation. So conventions are not separable
from the framework of institutional facts that effectives, by falling
under conventions, elaborate and help sustain.
6.3.2. Verdictives
Whereas effectives produce facts, verdictives are merely determina-
tions of fact. They have official and binding consequence, however,
and what they determine to be so is the case, as far as the institution is
concerned. Since what is determined may be a matter of natural rather
than institutional fact, we cannot say that verdictives make it the case
that such and such, only that they make it as if it were the case in the
sense that it is a fact for the institution in question. There is nothing
metaphysically mysterious about these as-if facts. Their status is epis-
temological, not ontological. To say that verdictives produce as-if facts
means, from a practical point of view, that as far as further institutional
processes and procedures are concerned, what has been determined to
be so is then acted upon as if it were so.
Verdictives are commonly determinations of natural fact. They settle
such questions as who killed Jones, whether Smith was tagged, and
whether Johnson signed his name to a certain document. Since such
natural facts are describable in institutional terms (that someone mur-
dered Jones, that Smith was tagged out, and that Johnson forfeited his
rights), verdictives are determinations of institutional fact as well. In
some cases, institutional fact, rather than natural fact, is directly at
issue. For example, there may be no dispute about what Brown did to
Jones, the only questions being whether he acted in his capacity as
officer of the law and whether he did so rightfully. In general, verdic-
tives serve the institutional purpose of settling issues in order that
institutional activity can carryon, proceeding from what has been de-
termined to be the case.
A verdictive would be merely a constative if it had no institutional
import of the sort described. However, verdictives do have this import
and can have it only in virtue of being mutually believed to have it. In
short, they must be conventional. Utterances that in other contexts
would be taken as mere constatives count, depending on the case, as
findings, rulings, certifications, etc., only because they are issued by
The Theory 116
the right person in the right situation (right, as conventionally deter-
mined).
It turns out, then, that effectives and verdictives must be conven-
tional. They are utterances that satisfy certain descriptions couched in
institutional terms, at least when the meaning of the conventional verb
is unpacked. As such, they are acts of either creating institutional facts
or of officially determining there to be certain facts, whether institu-
tional or natural. Either way, these utterances count for what they are
only in virtue of being mutually believed to so count; thus they must be
conventional.
6.4. PERFORMING CONVENTIONAL ILLOCUTIONARY ACTS
Both communicative and conventional illocutionary acts are utterances
issued with the intention that the utterance count as an act of a certain
sort. The means are different in the two cases, intention by means of
recognition of intention in one case and intention by convention in the
other. Nevertheless, they have in common the feature that the speak-
er's utterance, his saying that *( ...p ... ) by uttering e (or simply his
uttering e, in locution-specific cases), counts as F-ing that P only under
certain circumstances and never merely because of what e means. A
theory of illocutionary acts must explain how such utterances count,
when issued with certain intentions under certain circumstances, as
performances of illocutionary acts of certain sorts.
On our account communicative illocutionary acts are acts of ex-
pressing attitudes. To express an attitude is to R-intend the hearer to
take one's utterance as reason to think one has that attitude. For this
intention to be fulfilled, the hearer must identify it on the basis of what
is said together with mutual contextual beliefs, given the application of
the linguistic and the communicative presumptions. Whereas the LP
and the CP are general mutual beliefs within the linguistic community,
the mutual contextual beliefs are between the speaker and the hearer in
particular. Only because these mutual contextual beliefs are activated
(perhaps by the utterance itself) can the hearer identify, and can the
speaker reasonably expect him to identify, the speaker's illocutionary
intent. Only thus can the illocutionary act succeed.
On the other hand, in the case of conventional illocutionary acts, a
given utterance counts as an act of a certain sort just in case what is
said (or uttered, in some instances) and the conditions of utterance
Conventional Illocutionary Acts 117
meet the specifications of the relevant convention. The convention is a
matter of general mutual belief in the community or group involved. No
R-intention is required for performing a conventional illocutionary act,
and the speech act schema is inoperative. Instead, the speaker simply
intends that his utterance count as an act of a certain sort merely in
virtue of falling under the relevant convention. Now his intention in
this case is not to express a certain attitude but to affect institutional
affairs; his intention is fulfilled if his utterance meets the required
specifications. Insofar as those present must reckon that these specifi-
cations have indeed been met, the only relevant contextual belief
mutual among those present in particular, as opposed to the community
or group at large, is that the utterance and the conditions of utterance
meet these specifications. In most cases it should be clear to all con-
cerned whether or not these specifications are met.
Conventional illocutionary acts are not essentially communicative
and do not require R-intentions. Indeed, as Strawson notes (1964, 456),
there may be no particular persons to whom such utterances are di-
rected: uptake is not so much their point as their verification. The
hearer's role, even when he is the addressee rather than a mere on-
looker, is to recognize that the utterance falls under a certain conven-
tion and only incidentally to recognize that it is issued with a certain
communicative intention (if any). Although the speaker's intention is
not essentially audience-directed and although no communicative pre-
sumption is involved in conventional illocutionary acts, the speaker's
intention is nonetheless overt. As a member of the group or community
in which the convention obtains, he may be presumed to know this
convention and hence to intend his utterance to fall under it. Indeed, he
may expect and thus intend to be so presumed, but such an intention is
inessential to the success of his primary intention to be performing a
certain conventional illocutionary act.
In drawing a sharp distinction between conventional and communi-
cative illocutionary acts, we do not mean to imply that both sorts of
acts can't be performed in the same breath. Quite the contrary, one and
the same utterance can count as an act of both sorts. If a policeman
says to a person "You're under arrest," he is both arresting the person
and telling him (indirectly) that he has violated the law. It is common
for utterances to be both conventional and communicative illocution-
ary acts; when a speaker performs a conventional act, he is likely to
have a reason for doing it, and he may very well R-intend the hearer to
The Theory Conventional Illocutionary Acts 119

infer what that reason is. This common occurrence does not vitiate the may provide, the act is nullified if it is shown not to have been per-
distinction between the two types of acts, because they are issued with formed intentionally.
different sorts of intentions and succeed in different sorts of ways. Finally, conventional illocutionary acts, though resulting in institu-
Conventional illocutionary acts are identified by the speaker's utter- tional change, are not necessarily final. These changes can be reversed,
ance and the conditions of utterance, with R-intentions unnecessary the act that effected them rescinded. In the case of verdictives, the
and the SAS inoperative. It would seem, therefore, that they cannot be judgment of the umpire or the court may be challenged or appealed.
performed.nonliterally. For eX:ample, it seems that to fire someone the The umpire can, conceivably, change his mind, and the judge can be
speaker must say something that means 'You're fired.' But suppose overruled by a court of appeals. Whether created by effectives or de-
he says "George, remember that gold watch you've always wanted? termined by verdictives, institutional facts live by mutual belief and die
Guess what! Now you're getting it." Is this an indirect act of firing? by mutual belief.
We're inclined to say that it is not, but instead an indirect way of telling
George that he's fired (or involuntarily retired). In general, conven-
tional illocutionary acts must be performed literally, since the conven-
tions that govern them specify what must be said, not meant. However,
in locution-specific cases, it is irrelevant whether or not the utterance is
literal. All that matters is that the right words be uttered.
For a conventional illocutionary act to be successful, either by
changing an institutional state of affairs or by being an official determi-
nation of fact, the speaker's utterance and the conditions of utterance
must meet the specifications of the relevant convention. However, that
it meets these specifications is itself something that must be deter-
mined, so when there is a question whether a certain act was performed
and thereby whether a certain institutional change took place, the
speaker is not the sole judge. Hearers, perhaps positionally privileged
ones, may be consulted to verify that a certain utterance was genuinely
issued and that the circumstances were of the right sort. If there is a
dispute on this question subsequent to the time of utterance, special
procedures may be required to determine whether or not a certain
conventional illocutionary act had been duly performed.
One issue that may arise when it becomes necessary to determine
whether a certain act has been performed is the speaker's intention to
be performing that act. In some cases it does not matter, as with bid-
ding in the game of bridge. In general, however, the speaker is pre-
sumed to be intending to be performing the act that his utterance counts
as the performance of under the circumstances. Although it is possible
to find someone guilty while intending to acquit him (one might not
correct onself after having omitted the word "not") or to resign jok-
ingly (the boss seizes an opportunity of not having to fire an employee
when the latter jokingly says "I quit"), generally, as institutional rules

Chapter Seven Communication and
Convention
Effectives and verdictives are fundamentally different from communi-
cative illocutionary acts. They are implemented not by recognition of
intention but by convention. This is possible-and necessary-be-
cause effectives and verdictives essentially affect institutional states of
affairs and are only incidentally communicative. Now we have not
shown explicitly that communicative illocutionary acts are not con-
ventional. Austin held that all illocutionary acts are conventional, but
unfortunately, he did not specify what he meant by "conventional."
The closest he ever came was when he said that the use of a sentence
with a certain illocutionary force "may be said to be conventional in the
sense that at least it could be made explicit by the performative for-
mula" (1962, 103). This strangely qualified remark makes it evident that
whatever Austin meant by conventionality, it was something quite dif-
ferent from what we mean; it suggests that there is some special sense
in which explicit performative utterances are conventional. We con-
tend that communicative illocutionary acts are not conventional, in our
sense. We cannot show that they are not conventional in every sense,
however bizarre, but we argue that Austin's sense is no sense of "con-
ventional." Indeed, we argue that explicit performative utterances are
not conventional at all and that their special character can be ac-
counted for via the notion of illocutionary standardization. Finally, we
indicate one way in which all utterances are conventional-as locu-
tionary acts.
7.1. RULES AND CONVENTIONS
In order to evaluate claims to the effect that the use of language is
rule-governed behavior or that it is conventional, it is imperative that
Communication and Convention 121
the notions of rules and of conventions be distinguished. On our con-
ception conventions are not what people expect one another to do in
specified situations. Rather, they are actions that, if performed in cer-
tain situations, count as doing something else. This was reflected in our
definition, repeated here.
Convention: A (in C) is a convention for D-ing in G if and only if:
i. it is MB in G that whenever a member of G does A in C, he is D-ing,
and
ii. A in C counts as D-ing only because it is MB in G to count as such.
Rules, on the other hand, are socially expected forms of behavior.
Several types of rules are distinguished in the appendix, but for our
purposes it is enough to mention what they have in common. Indeed,
we will restrict ourselves to the central sort of case.
It is ambiguous to say that a rule is a socially expected form of
behavior. This can mean either that it is expected in a group G that
people will do A when C arises or that it is regarded in G that people
should do A when C arises. Our conception of rules reflects both the
anticipatory and the normative aspects of social expectations.
Rule: A (in C) is a rule in G if and only if:
i. the members of G do A in C,
ii. it is MB in G that (i), and
iii. it is MB in G that the members of G should do A in C.
Clause (i) has it that A is actually done, but this condition can be
relaxed to cover less central cases. It is difficult to imagine a situation
where clause (ii) could be true for long unless (i) were generally true.
On the other hand, (iii) could and often does obtain without (i) and (ii)
holding. A type of action that satisfies (i) alone is a statistical regularity;
one that satisfies both (i) and (ii) is a social regularity; and one that
satisfies (iii) is a social standard. The appendix may be consulted for
further explanation and for an argument justifying the requirement of
mutual belief in G, rather than merely shared belief.
Having distinguished rules from conventions, we can now make sev-
eral observations about rules and their application to illocutionary acts.
First, it should be plain that not every rule-governed act is conventional
and not every conventional act is rule-governed. Workers might be
required not to smoke, but there is nothing conventional about not
smoking. On the other hand, raising one's hand in a classroom conven-
tionally counts as an act of seeking recognition for purposes of speak-
The Theory 122
ing, but there is no rule requiring one to raise one's hand; if there is a
relevant rule here at all, it is that one should not speak out of tum and
not use other means than the designated one of getting attention (such
as jumping up and down or banging on the table). Although rules and
conventions are distinct, an action can be both rule-governed and con-
ventional. It might be a rule that spectators rise when the judge enters
the courtroom, and it is a convention that rising when the judge enters
counts as a show of respect. With the distinction between rules and
conventions in mind, we may say that rules can be observed or vio-
lated, whereas conventions can be merely applied or misapplied. Rules
specify things to be done, conventions ways of doing things. Clearly,
then, not every act performed in the context of an institutional practice
or procedure is conventional.
7.2. WHY COMMUNICATIVE ILLOCUTIONARY ACTS ARE NOT
CONVENTIONAL
Why are only effectives and verdictives conventional? To answer this
question we will run through the four categories of communicative
illocutionary acts and show why each cannot be conventional (except
in the irrelevant sense of being performed through conventional means,
that is, linguistically). It may seem especially plausible to regard com-
missives and acknowledgments as conventional. After all, commissives
create obligations, and acknowledgments are part of social rituals.
However, these facts do not make them conventional.
In arguing that only effectives and verdictives are conventional, we
imply that none of the other four categories of illocutionary acts over-
laps either of these two categories. Yet of the effective and verdictive
verbs listed in section 6.2, some have appeared before in other cate-
gories. For example, the verdictive "certify" was listed as a constative
(in particular, a confirmative), and the effective "bid" was included
under commissives. However, just as verbs like "inform" and "order"
cut across the illocutionary/perlocutionary distinction and verbs like
"tell" and "suggest" cut across communicative categories (being both
constative and directive), so certain verbs bridge the communicative/
conventional gap. We cannot expect ordinary usage to be sensitive to
every philosophical or other theoretical distinction, and so it should
come as no surprise that the same verb can name both a communicative
and a conventional illocutionary act type. Nothing in our formulation
of the distinction between the two kinds of acts precludes the possibil-
Communication and Convention 123
ity that one and the same utterance can be both. Although we argued
that conventional acts don't require R-intentions, surely they can be,
and indeed often are, issued with R-intentions: when a boss fires some-
one, he may be also telling the employee what he thinks of his per-
formance.
Now in considering whether communicative illocutionary acts can be
conventional, we must not trivialize the issue by arguing that since
communicative intentions are R-intentions and conventional inten-
tions are not, communicative acts can't be conventional. The issue is
whether the kinds of illocutionary effects produced communicatively
can also be produced conventionally. Communicative effects could be
achieved conventionally if certain conventions and rules existed. For
example, there could be a rule that people are never to speak nonliter-
ally or indirectly. Indeed, they could be required always to use the
explicit performative formulas "I state," "I request," and so on. No
inference along the lines of the speech act schema would then be re-
quired to identify the illocutionary act being performed. The performa-
tive verb would make the force (type of act) explicit, and the rest would
make the propositional content explicit. Finally, there would be a con-
vention stipulating that whenever someone says that he F-s that p, his
utterance counts as F-ing that p. Given a rule requiring sincerity, an
utterance would thereby obligate the speaker to have just those atti-
tudes that he would be expressing in communicatively F-ing that p.
Thus, in saying that he F -s that p, S would intend conventionally,
rather than by means of recognition of intention, his utterance to be
reason for H to think S has those attitudes. In this way, anyone could
express his attitudes conventionally. 1 Whether a person actually has
the attitudes he expresses would be another matter.
We have conceded that communicative illocutionary acts could be
performed conventionally if certain conventions existed. However, as
things are, can they be conventional? We will consider the four cate-
gories of communicative illocutionary acts individually.
Constatives
Ignoring the differences between the various types of constatives, we
may say, speaking generally, that they consist in expressing a certain
belief together with the intention that the hearer believe likewise. As
things are, this cannot be a matter of convention. Of course there are
conventional means for constating, such as using performatives like "I
report" and "I dispute," but it does not follow that using these means
The Theory 124
counts ipso facto as expressing a belief (and the corresponding inten-
tion). The hearer might have good reason to think that the speaker is
not being serious or literal in his utterance and conclude that no belief is
being expressed. Even using conventional means for constating, the
speaker must have the requisite R-intention, which is to be recognized
not only by what is said but also by supposing that the speaker is being
serious and literal in his utterance. Thus, conventions such as we have,
as opposed to those imagined, cannot supplant R-intentions in deter-
mining the identity of constative utterances.
Directives
Similar reasoning applies to directives. In general (considering the vari-
ous types), directives express the speaker's intention that the hearer act
because of S's desire or S's utterance itself. S succeeds communica-
tively if H recognizes this intention. Even if standardly directive forms
of words are used, such as the imperative mood or directive performa-
tives, S may not have the requisite intention. As in the case of con-
statives, there is no guarantee that S is being serious and literal in his
utterance and H may have good reason to think that he is not. So what
is uttered and what is said do not suffice to make an utterance directive.
Although effectives like licensing and prohibiting create rights or
obligations, it should not be thought that directives do likewise, at least
as illocutionary acts. At best, they create mutual beliefs between Sand
H about rights or obligations, and it is a moral question whether (or
when) they create rights and obligations. Even directives like authori-
zations and commands, which typically are issued within the frame-
work of some institution, create (institutional) rights and obligations
only if there are rules to that effect, for example, that people should do
what they are rightfully ordered to do. That such utterances are intra-
institutional does not make them conventional.
A wide range of institution-bound acts have the appearance of direc-
tives, acts such as legislating, licensing, and exempting. In some cases,
like ordering, prohibiting, and authorizing, they may even be marked
by directive verbs. They seem to have the force of directives inasmuch
as they require, prohibit, or permit certain types of action. However,
although they are directives in appearance, they are effectives in fact.
In particular, they are rule-making (or rule-unmaking) acts: they make
it the case in an institution or group that a certain action must be done,
must not be done, or may be done, whether by members in general or
by some specific category. Obviously, acts of legislating make laws,
Communication and Convention 125
which are a kind of rule, and legalizing something formerly prohibited
or rescinding legislation is to dissolve a rule. Acts of licensing and
exempting are not rule-making acts, for they apply to particular per-
sons and are, indeed, acts of permitting. Nevertheless, they have the
effect of removing a person from the scope of a rule that prohibits a
certain form of action, thereby changing his institutional status from
someone to whom the rule applies to someone to whom it does not.
Therefore, unlike communicative permissives these acts do not merely
provide a reason for the person to feel free to do something, they put
him in the position of being free to do it.
Commissives
Commissives are acts of undertaking obligations, but to undertake an
obligation is not automatically to create one, even if S uses a performa-
tive like "I promise." S's utterance may express his belief that an
obligation is thereby created, but that does not make the belief true
even if H shares the belief and it is mutual. That S is obligated to fulfill
his commitment is a moral question not answerable by the theory of
illocutionary acts. Of course, in institutional contexts it makes clear
sense to speak of obligations (which mayor may not be moral) that
exist in the institution as a matter of mutual belief. For commissives
issued in institutional contexts to create institutional obligations, there
must be institutional rules to the effect that one should honor one's
commitments, but this does not make them conventional. However,
formalized acts like taking an oath or signing a document are genuinely
conventional. These acts count not merely as undertaking an (institu-
tional) obligation but as creating one, and that they count as such is a
matter of convention, independent of any R-intention S may have.
Acknowledgments
In considering whether acknowledgments are conventional, we must
be especially careful not to be confused by the ambiguity of the term
conventional. Clearly such acts as greeting, thanking, and apologizing
observe social convention in some sense of the term. They are part of
everyday social rituals, they are expected on certain occasions, and
they can be performed with an air of perfunctory formality. But none of
this makes them conventional in the sense of falling under counts-as
rules. If they were conventional in this, the relevant sense, they would
succeed not in virtue of H's recognizing S's R-intention but by falling
under a convention.
The Theory 126
One reason that acknowledgments might seem to be conventional (in
the required sense), and not merely expected acts of social propriety
(an irrelevant sense), is built right into our definitions of them. Each
definition is disjunctive, and the second disjunct is of the form "S
expresses the intention that his utterance satisfy the social expectation
... and the intention that H take S's utterance as satisfying this expec-
tation." Acknowledgments that satisfy this disjunct are the perfunc-
tory, neither sincere nor insincere, gestures of everyday civility. The
social expectation referred to reflects a social rule to the effect that one
issue an acknowledgment when circumstances call for it. That ac-
knowledgments satisfy such rules does not make them conventional.
Even if the speaker is not expressing any feeling, as he would be if his
utterance satisfied the first disjunct of the definition of acknowledg-
ments, that it satisfies the second disjunct is still a matter of R-
intention. The obvious insincerity of a perfunctory acknowledgment
doesn't mean it isn't issued with an R-intention, namely, that the utter-
ance satisfy a social expectation.
Another feature of acknowledgments that makes them seem conven-
tional is that a distinctive locution, often of just one word, is associated
with each type: "Thank you," "Congratulations," "Hello." Does this
mean that they are conventional, indeed locution-specific? Although
each such form has a specific illocutionary use as a matter of linguistic
convention (of meaning), still these forms can be used nonliterally or
unseriously. Thus, an R-intention is required if an utterance of any of
these forms of words is to be a genuine acknowledgment. 2
The upshot of this argument is that in the absence of such conventions
as those in the fanciful situation described at the outset, communicative
intentions must be R-intentions-conventional intentions are no sub-
stitute. This is apparent from our analyses of the four categories of
communicative illocutionary acts. In each instance the speaker ex-
presses certain attitudes, sometimes rather complex but always as a
matter of R-intention. In the hypothetical situation in which there are
rules and conventions to give hearers reason so to take speakers' utter-
ances, R-intentions would be unnecessary. However, people could
speak only literally and people would be presumed to have the attitudes
they conventionally expressed. As things are, on the other hand, the
only conventional illocutionary acts are effectives and verdictives: up-
take is not so much their point as their verification. That a conventional
illocutionary act counts as an act of a certain sort is not a matter of
Communication and Convention 127
R-intention but of what convention it falls under. Thus, although in
practice a given utterance can be both communicative and conven-
tional and the same verb can designate acts of both sorts, the difference
between the two, in theory, is clear and distinct.
7.3. CONSTITUTIVE RULES AND COMMUNICATION
In our view an utterance counts as a communicative illocutionary act of
a certain sort as a matter of R-intention not convention. Successful
communication consists in the hearer's recognition of that intention by
way of an inference in the pattern of the SAS.
A rival view, which has gained considerable currency, is Searle's
(1969) theory of illocutionary acts based on the notion of constitutive
rules. Although only a detailed discussion would do full justice to the
theory, we wish to suggest that the underlying difficulty with this ap-
proach is that (a) Searle has given us insufficient reason to believe that
there are rules constituting types of communicative illocutionary acts
and (b) even if there were such rules, the most they could explain
would be the literal (and direct) performance of the illocutionary acts
with which they are associated.
Searle introduces the notion of constitutive rules not by definition
but by description and example. "Constitutive rules do not merely
regulate, they create or define new forms of behavior" (p. 33), such as
playing football or chess. To create the possibility of new forms of
behavior, constitutive rules take the form: X counts as Y in context C
(p. 35). This is essentially the form of what we call conventions. Since
Searle nowhere gives an explicit definition or a set of necessary and
sufficient conditions for the existence of a constitutive rule, as far as we
can tell, our notion of convention is congruent with Searle's notion of
constitutive rules (at least those that take the counts-as form 3 ).
How does Searle connect constitutive rules with illocutionary acts?
He writes, "The procedure which I shall follow is to state a set of
necessary and sufficient conditions for the performance of particular
kinds of speech acts and then extract from those conditions sets of
semantic rules for the use of the linguistic devices which mark the
utterances as speech acts of those kinds" (p. 22). Notice that Searle
refers to the relevant constitutive rules as semantic rules; yet nowhere
does he give us reason to suppose that semantic rules, at least as con-
strued by linguists, qualify as constitutive rules (perhaps only those that
"create" illocutionary acts so qualify). More serious is his failure to
The Theory 128
explain why it should be that to "extract from those conditions sets of
semantic rules" will give us the slightest reason to believe in the exis-
tence of these rules. Indeed Searle does not explicitly say how the rules
and conditions are related. Given other things he says, it is natural to
suppose Searle holds the following:
Rule-Condition Connection (R-C):
If R are the rules for promising and C are the conditions for promising,
then in uttering e, S obeys R iff in uttering e, S satisfies C. (We use e
(for expression) instead of Searle's symbol T.)
Searle then (1969, 54-61) turns to the analysis of promising as his
"initial quarry." But he does not really deal with promising per se, only
with nondefectively promising in literally uttering a sentence e to a
hearer H. We will see that this qualification makes a difference. Searle
presents nine singly necessary and jointly sufficient conditions for
promising (pp. 57-61), which we label and group as follows (phrasing
and all but the last two labels are Searle's):
(a) Input-output conditions:
1. Normal input and output conditions obtain.
(b) Propositional content conditions:
2. S expresses the proposition that p.
3. In expressing that p, S predicates a future act A of S.
(c) Preparatory conditions:
4. H would prefer S' s doing A to his not doing A, and S believes H
would prefer his doing A to his not doing A.
5. It is not obvious to both Sand H that S will do A in the normal
course of events.
(d) Sincerity condition:
6. S intends to do A.
(e) Essential condition:
7. S intends that the utterance of e will place him under an obliga-
tion to do A.
(f) Grice condition:
8. S intends (i-I) to produce in H the knowledge (K) that the utter-
ance of e is to count as placing S under an obligation to do A. S
intends to produce K by means of the recognition of i-I, and he
intends i-I to be recognized in virtue of (by means of) H's knowl-
edge of the meaning of e.
Communication and Convention 129
(g) Literal condition:
9. The semantical rules of the dialect spoken by Sand H are such
that e is correctly and sincerely uttered if and only if conditions 1-8
obtain.
From these conditions Searle (p. 63) extracts the following rules for the
indicator of illocutionary force, in this case Pr for promising (phrasing
and labels are Searle's):
R(b) Propositional content rule:
Rule 1. Pr is to be uttered only in the context of a sentence (or larger
stretch of discourse) e, the utterance of which predicates some future
act A of the speaker S.
R(c) Preparatory rules:
Rule 2. Pr is to be uttered only if the hearer H would prefer S' s doing A
to his not doing A, and S believes H would prefer S' s doing A to his not
doingA.
Rule 3. Pr is to be uttered only if it is not obvious to both Sand H that S
will do A in the normal course of events.
R( d) Sincerity rule:
Rule 4. Pr is to be uttered only if S intends to do A.
R(e) Essential rule:
Rule 5. The utterance of Pr counts as the undertaking of an obligation
to do A.
Clearly these rules do not accord with R-C. One could utter e in ac-
cordance with these rules and not literally and nondefectively promise.
What are missing are rule analogs to conditions (a), (f), and (g). In
saying that these conditions "apply generally to all kinds of normal
illocutionary acts and are not peculiar to promising" (p. 62), Searle
seems to suggest that there are some general rules that, if followed, will
guarantee that these conditions are met. But this remark and suggestion
seem false.
The literal condition (g) is not necessary for the performance of
illocutionary acts. The one-word sentence "Yes" can be literally ut-
tered in the making of a promise, yet it is not governed by semantic
rules that guarantee that conditions (a) -(f) obtain. Indeed, the only
sort of sentence that might satisfy condition (g) is an unambiguous 4
explicit performative, whose meaning determines and exhausts its
force. Besides artificially restricting his analysis and making it much
The Theory 130 Communication and Convention 131

less interesting, Searle makes his analysis easily falsifiable, since these performances of illocutionary acts. The reason for this is that the rules
restrictions are not built into the sentence e in the analysandum. for a given type of illocutionary act are presented as rules for using the
The general problem is how to write rules that accord with R-C. associated illocutionary force indicating device. Thus, for example, the
What happens when one actually tries to formulate them as analogs of rules for promising are presented as rules for using Pro Obviously, Pr
the conditions that they are to guarantee the satisfaction of? First, does not need to occur in a nonliteral or indirect promise, and thus the
corresponding to conditions (a), (f), and (g) we would have the rules: constitutive rules for promising cannot be invoked to explain promises
made nonliterally or indirectly.
R(a) Input-output rule: Pr is to be uttered by S (to H) only if normal
Our intention/inference approach to communicative illocutionary
input and output conditions obtain.
acts provides enough room between meaning and illocutionary force to
R(f) Grice rule: Pr is to be uttered by S (to H) only if condition (f) handle nonliteral and indirect, as well as literal, performances of illo-
obtains. cutionary acts. Whereas Searle uses his conditions for performing illo-
R(g) Literal rule: Pr is to be uttered by S (to H) only if the semantic cutionary acts to attempt to extract constitutive rules, for us they are
rules of the dialect spoken by Sand H are such that Pr is correctly and nothing more than conditions for reasonably expecting one's communi-
sincerely uttered iff conditions (a) -(f) obtain. cative illocutionary intentions to be recognized, namely, in accordance
with the SAS. Satisfaction of the input-output conditions enables the
Recall that R(g) must (in accordance with R-C) guarantee that condi-
hearer to identify the utterance act. Satisfaction of the propositional
tion (g) is satisfied if R(g) is followed, and that R(g) is a semantic rule
content and the preparatory conditions (S's predicating, literally or
governing e (containing Pr). There is a problem here with R(g). It will
otherwise, a future act of S that is mutually believed H would prefer
be satisfied if the semantic rules of the dialect are such that conditions
and S would not otherwise do) enables H to delimit the types of illocu-
(a) -(f) are satisfied. But of course these conditions will be satisfied if
tionary acts that could be performed in S's utterance. And satisfaction
rules R(a)-R(f) are satisfied. So R(g) is redundant.
of the essential condition enables H to identify just what that illocu-
The trouble is that Searle has gone too far in formulating as condition
tionary act is. For us the sincerity condition need not be met for the act
(g) what he takes to be the noncontingent relation between what e
to succeed communicatively, and the Grice and the literal conditions
means and what S meant in uttering it. It is not necessary for literally
are inapplicable to our formulation.
and nondefectively promising in uttering e that the semantic rules of e
be such that e is correctly uttered iff one is thereby promising (ambigu-
7.4. PERFORMATlVES AND CONVENTION
ity and synonymy seem to falsify this in each directionS); therefore,
why not drop R(g) and let R(f) carry the weight? Presumably because
We noted at the beginning of this chapter that Austin held that all
R(f) does not yet rule out nonliteral utterances of e (like sarcasm or
illocutionary acts are conventional but that he neglected to explain
metaphor). But why should literalness be a part of the analysis of
what he meant by "conventional." An illocutionary act is, he says, "an
promising? Surely there are less radical ways of stating what is true in
act done as conforming to a convention" (1962, 105) and "is constituted
the observation that one can't mean just anything by uttering just any-
not by intention or by fact, essentially, but by convention (which is, of
thing under any conditions. 6 A more natural and general approach
course, a fact)" (p. 127). The only clue to what he actually meant by
would be to define the notion of promising in uttering a sentence e and
"conventional" is his bewildering remark that the illocutionary force of
then define a special case where the utterance of e with its full conven-
an utterance "may be said to be conventional in the sense that at least it
tional force is sufficient for promising.
could be made explicit by the performative formula" (p. 103). How-
The fundamental problem for Searle's account of illocutionary acts
ever, as Strawson observes (1964a, 445), there seems to be no such
in terms of constitutive rules, apart from the question of the existence
sense of "being conventional," and "if this is a sense of anything to the
of these rules (those tied specifically to illocutionary acts), is how to
purpose, it is a sense of 'being capable of being conventional'." This
specify the rules in such a way as to allow for nonliteral or indirect
suggests that explicit performatives are conventional.
 The Theory 132
Strawson seems to hold this view. He categorically denies that all
illocutionary acts are conventional: "Some illocutionary acts are con-
ventional; others are not" (p. 445). Indeed, it was Strawson who, later
in this paper, introduced the idea that the nonconventional acts are
performed with reflexive (Gricean) intentions. Why does he think that
explicit performatives are conventional? Whereas communicative illo-
cutionary acts in general are issued with an intention to produce a
certain response "by means of recognition on the part of the audience
to produce that response, this recognition to serve as part of the reason
that the audience has for its response" (p. 450), the "peculiar logical
character" of performatives is "to make explicit the type of communi-
cation intention with which the speaker speaks, the type of force which
the utterance has" (p. 451). Apparently, Strawson thinks that if the
form of words makes explicit the illocutionary force, no reflexive in-
tention is necessary or need be ascribed by the hearer to the speaker.
Instead, it can be read off of what is said, that is, "illocutionary force is
exhausted by meaning" (p. 456).
Our reply to this view should come as no surprise. Even where the
speaker is being not only literal but fully explicit about the force of his
utterance (as when he uses the explicit performative formula) still he
must R-intend to be taken as speaking literally and explicitly and must
be so taken. Meaning never exhausts illocutionary force, even when
the speaker is doing precisely what he says he is doing and nothing else.
The meaning may fully specify what in fact the speaker is doing, but it
does not determine that this is what he is doing. Thus, just as with
literal illocutionary acts in general, an inference in the pattern of the
SAS is required even for explicit performative utterances. And the
speaker must R-intend the hearer to make such an inference. We will
leave our reply to Austin and Strawson as it stands until we take up the
subject of illocutionary acts and linguistic devices in chapter 10.
7.5. WHY LOCUTIONARY ACTS ARE NOT FULLY
CONVENTIONAL
Communicative illocutionary acts are not conventional qua illocution-
ary acts. Yet there seems to be an important, though obvious, respect
in which all utterances in a language are conventional. It is worth
spelling out in what respects utterances are and are not conventional.
Strawson (1964a, 442), in examining Austin's view that all illocution-
ary acts are conventional, contrasts this with the uncontroversial view
Communication and Convention 133
that "we must refer. . . to linguistic conventions to determine what
locutionary act has been performed in the making of an utterance, to
determine what the meaning of the utterance is." Schiffer (1972, 155)
describes utterance types as "conventional means for communicat-
ing." Without saying just what conventional means are,7 he does indi-
cate that they take the form "if one does X, then such-and-such will be
the case" and that this is a matter of mutual knowledge (his analog to
our notion of mutu~l belief). Since this parallels our notion of conven-
tional action, we can use that notion to state to what extent an utter-
ance in a language is conventional qua locutionary act.
It is natural to think of language as conventional in the sense that the
meaning of its words is conventional. In our terms, what words mean is
what we mutually believe them to mean. To be sure, no one knows
what every word means, so allowances must be made for words not in
the common parlance, such as technical terms. Even then at least it is
mutually believed in the linguistic community that such terms mean
whatever the relevant experts mean by them. In any case, there are
thousands of words in the core vocabulary of the vernacular, and virtu-
ally everyone shares the mutual beliefs about their meanings.
People's (conscious) knowledge of the principles that determine how
words (or formatives) combine to form determinately meaningful sen-
tences is less explicit than their knowledge of the vocabulary. Members
of the general public are not linguists or grammarians, but they do
employ the linguistic presumption: that they share the language they
use and that sentences as well as words mean more or less the same for
all. They do not have explicit mutual beliefs about every particular
sentence but, given the LP, it is at least the case that whenever some-
one utters a sentence to someone else, they mutually believe that the
sentence e means such and such. Although it is a matter of convention
that e means ... , uttering e conventionally counts as saying that
*( .. .p .. . ) only when the meaning of e determines what S says (provided
S is saying something in uttering e).
Since locutionary acts are conventional only in this way, what a
speaker says, as opposed to what he does in saying it, is a matter of
R-intention only as regards resolving ambiguity and determining in-
determinate references. Just as a conventional illocutionary act is
something for the hearer to identify on the basis of his knowledge of the
relevant convention, so a locutionary act, insofar as it is determined by
what the utterance e means, is identified on the basis of H' s knowledge
of the language. Up to ambiguity, H does not have to figure out what S
The Theory 134

means by e by inferring that S has certain intentions. So whereas Scan Part Two Issues
deny that he has the illocutionary intent H attributes to him - by saying
he was not serious or was not speaking literally-S has no deniability
regarding what he says, at least as delimited by the meaning of e. One
can admit to a malapropism or plead slip of the tongue, but this only
shows that one can mean (to say) one thing and say another. To under-
stand what a speaker is doing in uttering something, it is of course
relevant to know that his utterance was intentional. Nevertheless,
given what was uttered, the identity of what was said is not a matter of
intention - it is determined, except for ambiguity and indefinite refer-
ence, by linguistic convention. These two factors keep most locution-
ary acts from being fully conventional.
Chapter Eight Locutionary Acts:
Philosophical and Linguistic
Issues

In part I we often used notions of meaning and occasionally notions of

presupposition and implication without saying very much about them.
We said enough for the purposes of the speech act schema, but then the
schema is designed to reflect ordinary patterns of inference by normal
hearers in response to speakers speaking normally. For this purpose we
could not stray very far from ordinary concepts, because the normal
speaker is neither a philosopher nor a linguist and so no philosophical
nor linguistic analysis should be imputed to him.1 Philosophers and
linguists, though, can raise certain conceptual and linguistic questions
about lines in (and aspects of) the schema.
Our first topic will be linguistic meaning. When we say that an ex-
pression has a certain meaning, what are we specifying and how is it to
be specified? More fundamentally, what is it for an expression to have
meaning? Then there is the topic of speaker meaning, of what a speaker
means by an expression. For our purposes the concept of operative
meaning, as explained in chapter 2, will suffice, although Grice and
Schiffer have introduced further notions of what a speaker means.
While operative meaning, together with reference, determines what is
said, it seems relevant also to what is presupposed and to what is
implied. Various notions of presupposition (semantic as well as prag-
matic) have been proposed, and Grice has offered a systematic account
of implication. Our general concern will be not only to unravel these
notions, but also to ascertain their place in our overall account of
linguistic communication. We will argue that some of these notions are
already incorporated in our account, while others are theoretically su-
perfluous. The reader should be warned that to keep the length of this
chapter within reasonable bounds, we must assume some familiarity
with the cited literature.
Issues 138
8.1. LINGUISTIC MEANING
In general the linguistic meaning of an expression is simply the meaning
or meanings of that expression in some linguistic system. This rough
characterization can be clarified in a number of ways, in particular, by
saying more about the format for specifying linguistic meaning and the
conditions for correctly attributing linguistic meaning.
At L2 of the SAS H identifies the operative meaning of e, so H must
be able to represent that reading of e to himself. To do this H must have
some system of (semantic) representation for specifying meaning. What
it is, what it looks like, we don't know, though not from lack of sugges-
tions. The literature contains proposals concerning the nature of se-
mantic representation ranging from (1) the language L itself (Evans and
McDowell 1976, Introduction), through (2) a regimented and aug-
mented version of L (Harman 1973; Davidson 1967, 1970), (3) an inten-
sionallogic (Lewis 1969, 1970; Montague 1974), (4) special systems of
conceptual notation (Katz 1972; Bierwisch 1970), to (5) a language of
thought (Fodor 1975; Miller and Johnson-Laird 1976). Even if we rule
out some of these as psychologically implausible (for example, present
intensional logics-see Partee 1977), the options are still widely di-
verse.
We reject view (1), according to which the semantics of sentences in
L can be represented to H by those very sentences of L, simply
because of the ambiguity (syntactic or lexical) of most sentences.
Operative meaning specifications identify the result of contextual dis-
ambiguation and so cannot themselves be ambiguous. Although it is
possible that for every n-way ambiguous sentence of English there are
n unambiguous English sentences each of which means just the same
thing as one reading of the original sentence, we think it most unlikely.
And we know of no evidence that hearers actually come up with such
sentences to represent operative meaning. Concerning view (2), some
regimented form of L, with no syntactic or lexical ambiguity, 2 avoids
the problems raised by ambiguity, but it renders hopelessly mysterious
how a person could ever learn the language in the first place or translate
between languages. This is no conclusive objection, for maybe these
are indeed mysteries. But we would prefer to trade them in for other
problems, such as motivating the special systems of (4), or for other
mysteries, such as the nature of a language of thought (5).
Many systems of semantic representation are compatible with the
schema, which demands only psychological plausibility and nonambi-
Philosophical and Linguistic Issues 139
guity in specifications of meaning. In phi~osophy .c~rrently the mo~t
popular approach is Davidson's (1970) project of gIvmg a truth-defim-
tion for a natural language, but we know of no way (and suspect there is
none) to make it psychologically applicable. In linguistics and psychol-
ogy the most popular approach seems to be decompositional, along the
lines of Katz (1972), Bierwisch (1970), and most versions of generative
semantics. In the spirit of our approach to communication in general,
we will tentatively endorse a more inferential perspective on seman-
tics, but first let us give some reasons for suspecting the decomposi-
tional approach.
8.1.1. Decompositional Theories
We consider a semantic theory to be decompositional just in case it
represents the meaning of a syntactically (or morphologically) un-
structured item as being composed of more than one semantic element,
that is, the semantic representation is complex. The classical statement
of such a semantics was Katz and Fodor (1963). The theory has under-
gone substantial revision and elaboration, becoming considerably more
sophisticated (see Katz 1972, 1977c, or Bierwisch 1970) while retaining
its decompositional character. The same is true of its once-bruited
alternative, generative semantics. On each of these theories a word like
"bachelor" or "chase" or "kill" is represented as having internal
semantic structure in the sense that parts of the semantic representa-
tion of each of these words can appear in the semantic representation of
other words and so represent the same semantic contribution to the
meaning of those words. Put another way, a decompositional semantics
extends compositionality to the internal structure of lexical items. In
effect the claim is that much the same representational machinery in
semantics can be used for syntactically structured and unstructured
expressions alike. Behind this claim is the idea that it is somewhat an
historical accident which semantic representations (or meanings) be-
come associated with a single word in a language and which ones get
associated with compound words or phrases. For instance, English has
the phrases "to intentionally kill oneself" and "to commit suicide," but
no single word to express this notion. According to a decompositional
semantics, this peculiarity should be viewed as only an historical acci-
dent, on a par with the fact that English has the single word "thumb"
whereas Japanese uses the equivalent of "mother finger."
The common goals of semantic theories, at least in linguistics (see

Issues 140
Janet Fodor 1977), are to attribute to each well-formed expression in
the language the correct range of semantic properties (meaningful,
meaningless, redundant, analytic, contradictory, ambiguous, and so
on) and relations (entails, contradicts, is synonymous with, is similar in
meaning to, presupposes, and so on), though one need not subscribe to
all these goals. A decompositional semantics usually is combined with a
pair of additional claims to the effect that (i) the representation of the
semantics of an expression defines the expression (gives necessary and
sufficient conditions for its correct application) and (ii) the attribution
of the range of semantic properties and relations to an expression is
mediated by the representation of its meaning either as given in the
definition (for syntactically/morphologically unstructured lexical items
and idioms) or as the result of compositional operations on such defini-
tions. Thus, on a semantics like Katz's the meaning, or meanings, of a
sentence such as "He is a bachelor" would be represented in the theory
by a formula! (or set of formulas) in the system of semantic represen-
tation such that! is the result of general principles of composition
applying to the definitions of the constituent words and their grammati-
cal relations, whereby! forms the basis for attributing the sentence's
semantic properties and relations. And! does this by means of a system
of formal theoretical definitions. For instance, being meaningful (as
opposed to meaningless) might be defined as having a semantic rep-
resentation in the theory, ambiguity can be defined in terms of receiv-
ing multiple semantic representations, synonymy in terms of receiving
the same semantic representation(s) in the theory, and so on. Such--
definitions are theoretical in that they are stated within the vocabulary
of the theory; they are formal in the sense that the correctness of their
application to cases can be determined by inspecting the form or shape
of the semantic representation.
Although we will return to these notions later, what we have said is
sufficient to see how one might motivate decompositional semantics.
Probably the central form of linguistic argument in favor of lexical
decomposition is simplicity of predictions concerning various seman-
tic properties and relations. 3 Just as transformational rules capture
generalizations missed by phrase structure grammars (Chomsky 1957;
Akmajian and Heny 1975, ch. 3), so it could be claimed that systems of
semantics without decomposition fail to reflect certain generalizations
concerning semantic properties and relations (see Akmajian, Demers,
and Hamish 1979, ch. 11).
Philosophical and Linguistic Issues 141
The price of such theoretical elegance is a concomitant strengthening
of constraints on the adequacy of a semantic description of a particular
language. Decomposition demands definitions of all meaningful words,
definitions yielding necessary and sufficient conditions of correct ap-
plication, as well as formal identity of synonyms and nonidentity of
nonsynonyms. Some theorists have found such constraints highly de-
sirable in principle but too strong in fact (see Fodor, Bever, and Garrett
1975, 209-214; Fodor, Fodor, and Garrett 1975, sec. 3; J. D. Fodor
1977, ch. 5.1, 5.6; J. A. Fodor 1975,147-156). The counterargument
proceeds in three stages.
First, the decompositional theory is softened up by arguing that de-
composition and theoretical definition are not sufficient machinery to
make all relevant semantic predictions. For instance, to predict that "is
a male mother" is contradictory, one needs a way of inferring that if
something expresses FEMALE, then it expresses NOT-MALE, so that the
representation of "is a male mother" could be shown to comprise
MALE(X) & NOT-MALE(X) and so be contradictory. To accommodate
such facts, Katz (1966, 1972) and others (Bierwisch 1970) incorporate
into the semantic component a set of redundancy rules, which mark the
first piece of semantic machinery supplementing the devices of decom-
position and theoretical definitions.
The second stage is to claim that the general requirement of for-
mulating decompositional definitions is too strong. Some relations can-
not be plausibly accounted for in these terms, for instance, the relations
between:
(1) necessary: not possible not
some x is: not every x is not
p and q: not (not-p or not-q)
If it is proposed that the right side define the left side, then what does
one do with their duals, which have equal claim to definitional status?
(2) possible: not necessary not
every x is: not some x is not
p or q: not (not-p and not-q)
Furthermore, sometimes only necessary conditions seem to be possi-
ble:
(3) green: COLORED
Issues 142
To get a sufficient condition one would have to fill in the ellipsis, but
what could so define "green" and not entail COLORED?
(4) COLORED & ... : green
In other cases what are offered as definitions are plausibly only neces-
sary conditions:
(5) kill: CAUSE (BECOME (NOT-ALIVE
Finally, there are types of expressions such as proper names, natural
kind terms, various particles, and perhaps performative verbs which
might not have necessary conditions of the sort required by decomposi-
tional theories. (See J. Katz 1975, 1977a,b,c, and references therein for
discussion of these kinds of cases and objections.)
The third stage of the case against decompositionality is to argue that a
semantic theory could not be a mixture of redundancy (or inference) rules
and decompositional definitions. So far the main complaint against
mixed theories has been that they provide no principled reason for
making a semantic prediction on the basis of decomposition as opposed
to inference (see Fodor, Fodor, and Garrett 1975, 522). However, there
is no reason at present to suppose a decompositionalist could not come
up with such a principle or principles (see G. Lakoff 1970, sec. 7). But if
principles could be found for drawing such a distinction, is it the sort of
distinction one wants? What is the difference, semantically, between
being necessitated by the inference rules
(6) father ~ MALE,
green ~ COLORED,
and being necessitated by the dictionary definitions,
(7) father: MALE & ... ,
green: COLORED & ... ?
If there is no difference in the semantic property being predicted,
should there be a difference in the semantic machinery that predicts it?
On the other hand, if a principle cannot be found for allocating predic-
tions to lexical definitions rather than inference rules, a mixed theory
would be unacceptable. But since the demand for decomposition is
sometimes too strong, inference rules would be the preferable single
piece of machinery to use (supplemented by definitions of semantic
properties and relations that any theory needs). Can an inferential the-
Philosophical and Linguistic Issues 143
ory be worked out? The matter is currently under dispute and much
remains to be done, but we can sketch the outlines of such an alterna-
tive and say how it fits the SAS.
S.1.2. Inferential Theories
One of the main attractions of an inferential theory for practicing
semanticists is the modesty of its ambitions. Instead of seeking defini-
tions as the basis for semantic description, it seeks entailments, and
these are generally easier to come by than synonyms. Such a semantics
has the virtue of making us preach what we practice. In the (rare?)
cases where one finds fairly uncontroversial sufficient as well as neces-
sary conditions, the inference-rules approach posits a two-way infer-
ence. Synonymy would be defined perhaps as 'mutual inferability,'
contradiction as 'entailing F and not-F,' and so on. Before attempting to
assess the range of semantic properties and relations that could plausi-
bly be covered by a theory of this sort, we should see what the specific
advantages of this approach might be. 4
Fodor (1975) sees at least three (nonpsychological) advantages to
inference rules over lexical decomposition. If a theory contains just
inference rules (and theoretical definitions), then, first, it would not
"posit a sharp distinction between the logical and nonlogical vocabu-
lary . . . the logical behavior of 'bachelor' is not, on this view, treated
fundamentally differently from the logical behavior of 'and' . . . the
entailments they engender are determined by the inference rules under
which they fall" (1975, 150). Second, although definition is a symmetri-
cal relation, entailment is not, and "there is no reason why, on the
present account, analyticity must rest upon symmetrical relations.
Some rules of inference go one way, other rules of inference go both
ways. There is nothing special about the latter" (1975, 152). Finally,
theories having both decomposition and inference seem to "posit an
unwarranted distinction in kind between formulae true by virtue of
definitions and certain other kinds of 'analyticity'" (1975, 148). In fact,
mixed theories fail to provide "principled grounds for claiming that the
two relations [of definition and entailment] have anything in common at
all" (1975, 149).
Although these may be advantages, they are not decisive. The sec-
ond argument is simply an endorsement of inferential semantics, and it
is not all that clear what the first and third considerations amount to.
Why are these distinctions unwarranted? Fodor does not say. It is
Issues 144
tempting to try to strengthen the point by arguing that homogeneity of
the entailment relation (or the univocality of "analyticity") requires a
single underlying mechanism to account for the pertinent semantic
facts. But such a move is in danger of proving too much. Applied to
ambiguity, it would establish the conclusion that a grammar should not
make use of different devices such as lexical entries, rewrite rules, and
transformations to account for lexical, surface, and underlying ambigu-
ities- surely a conclusion not warranted on any independent grounds.
Since one of the best ways of refuting a theory is providing a superior
alternative, we can ask what a semantic theory that used inference
rules as the basic semantic device would look like. According to Fodor,
Bever, and Garrett (1974, 184) it would have to contain (1) a set of
semantic representations (equivalent to a language), (2) a set of rules of
assignment which assign semantic representations to sentences, and (3)
a set of inference rules applying mechanically to the semantic rep-
resentations to determine entailments. Since the inference rules apply
to semantic representations, these representations must be rich enough
to determine (via the inference rules) all entailments of the expression
assigned that representation. It is thus appropriate to call such a rep-
resentation the logicalform of that sentence (in this system). The set of
entailments can be called the logicalforce of the sentence. Semantics,
in this conception, has as one of its basic tasks the specification of the
logical force of every expression in the language having such force.
What does the system of semantic representations look like? What
are the rules of assignment? What are the rules of inference? These are
matters for empirical investigation, but they are also partly matters for
decision. For instance, one must decide whether one's semantics is also
to be interpreted psychologically as an account of what a person is
intending to communicate when using a sentence literally. If one de-
mands psychological reality of one's semantic machinery, the con-
straints on it are increased appreciably (see chapter 11; also Fodor
1975; Harnish 1977c). Apart from questions of psychological reality,
what can be said about the nature of the inference rules? Since in the
Fodor, Bever, and Garrett (1975) account, inference rules are defined
over semantic representations, the exact nature of these inference rules
cannot be determined independently of the investigation into the nature
of semantic representation. But because it is plausible to believe that
the result of applying one inference rule can be in the domain of ap-
plication of a different rule, such rules of inference seem to determine
sets of pairs of semantic representations, (F1, F 2>' each pair having the
Philosophical and Linguistic Issues 145
characteristic that the sentence expressing F 1 entails (on that reading)
the sentence expressing F 2 If the theory is to be finite (or finitely
representable) it will not do for rules of inference to be such pairs: there
is a potential infinity of them. Inference rules must therefore collect
together classes of semantic representations. Traditionally this is done
in terms of crucial logical words that occur, in the sentences to be
represented. Sentences of the grammatical form "S and S" are rep-
resented by a wff of the form" ... & __ "; sentences of the form" S or
S" are represented with a wff of the form" ... v __ "; and so on. An
inference rule defined over such representations might have the form: 5
(8) From a wfI of the form " ... &--" infer a wfI of the form " ... " and
infer a wfI of the form " __ ."
The logical force of a sentence would be determined by applying the
relevant inference rules to its semantic representation.
Having gone this far down the road from decomposition in account-
ing for entailments, one might wonder whether there is reason for
retaining a level of semantic representation at all. Why not just have
rules legitimating inferences from one sentence (under a description) to
others? What else is needed in a semantics in which the basic semantic
notion to be captured is entailment? Two considerations in favor of
semantic representations come to mind. First, a semantic representa-
tion can function as an object of understanding, belief, confirmation,
and so on, and an account of understanding a sentence, belief, and so
on, can plausibly be given in terms of recovering or computing this
representation (see Fodor, Bever, and Garrett 1975, 374-384; J. A.
Fodor 1975). Since questions about the nature of a representation sys-
tem functioning in this capacity will be taken up in chapter 11, we will
not pursue this issue here.
Second, the existence of a semantic representation allows us to make
sense of the recurrent observation that grammatical form is misleading
as to logical form; without logical forms it might be thought difficult to
explain such disparities. Philosophers who make the observation rarely
say exactly how grammatical form is misleading as to logical form.
Strawson, however, has made such an attempt:
We might reasonably say that the verbal form of a statement was (at
least potentially) misleading as to its logical form in the following cir-
cumstances: (1) the sentence used to make it has a certain verbal pat-
tern in common with a great many other sentences; (2) most, or a great
number, of the statements made are analogous to one another in a
Issues 146 Philosophical and Linguistic Issues 147

certain formal respect; (3) the statement in question is not analogous to PM1 S PM2 S

~VP A
these statements in that formal respect. (1952, 51)
There is something importantly right here, despite the obscurities. We NP NP VP

I~NP
will assume that sentences have a logical form and will adapt Straw-
son's remarks to that end.
The grammatical form of s is misleading if it makes s appear to have N Copula
IA
N Copula NP
particular logical form F that it does not really have. 6 That is,
~ /\
Ir1
1. s has the grammatical form T; Det N

2. s appears, in virtue of its grammatical form T, to have the logical
form F;
3. s does not have the logical form F.
More needs to be said about the phrase" appears in virtue of." Taking a
clue from Strawson that most, or a great many, sentences with gram-
matical form T have the logical form F, then grammatical misleading-
ness comes to this:
1. s has a grammatical form T;
2. Most, or a great many, sentences with the grammatical form of s
have the logical form F;
3. s does not have the logical form F.
According to this account (potential) misleadingness is the result of
(potential) overgeneralization.
Philosophers who subscribe to the doctrine of logical form seem to
take a rather traditional view of grammar and grammatical form. How-
ever, since contemporary grammars reconstruct the traditional notion
of grammatical structure in terms of phrase-markers, to say of a sen-
tence that it has a particular grammatical form Tj is just to say that an
optimal grammar would assign some particular phrase-marker to that
sentence. Thus,
II
is is
Figure 8.1
sense of the notion of misleadingness without appealing to some formal
system? We can, and something closer to such an explication of mis-
leadingness lies behind many of the traditional examples of how gram-
matical form is misleading as to logical form. Let St, S2, ,Sm, be a
nonredundant enumeration of all and only the sentences of the natural
language L under consideration. Let PM l , PM 2 , , PM n , be a non-
redundant enumeration of all and only the phrase-markers for the sen-
tences of that language. This can be done with a phrase structure
grammar. As Chomsky remarks, "The natural mechanism for generat-
ing Phrase-markers ... is a system of rewriting rules" (1965, 66). Then
let PMk[s] be the equivalence class of sentences that have the phrase-
marker PMk assigned to s. Finally, suppose there to be an enumeration
of grammatical inference rules, Rh,j, Rj,k, ... ,R7,m, each of the form,
(9) R~,z: From a sentence assigned the phrase-marker PMy infer a sen-
tence assigned the phrase-marker PMz .

The grammaticalfarm of a sentence s is misleading as to its lagicalfarm
iff:
i. s is assigned the phrase-marker PMj by some (correct) grammar,
ii. Most, or at least a great many, sentences assigned the phrase-marker
PMj have the logical form Fj,
iii. s does not have the logical form Fj.
So far our explication of the misleadingness of grammatical form has
utilized the notion of a semantic or logical representation. Can we make
For example, consider the phrase-markers in figure 8.1, which would
probably occur in an enumeration of the phrase-markers of English.
And consider the grammatical inference rule,
(10) R1,2: From a sentence of the form PM l infer a sentence of the form
PM 2 -identical branches on both trees to receive the same words.
For instance, from "John is a strong boy" infer "John is a boy." We
can now say what it is for a sentence, in virtue of its grammatical form,
to be misleading as to its logical form:
 I
I
Issues 148
The grammaticalform of a sentence s is (at least potentially) misleading
as to its logical form iff:
i. s is assigned the phrase-marker PMi by a correct grammar,
ii. There is a grammatical inference rule Rlj that is valid for most or a
great number of PMi[S] ,
iii. R1,j is not valid for s.
By this definition, the two phrase-markers in figure 8-1, and inference
rule (10), we get the result that the sentence "John is a converted
heathen" is misleading as to logical form, since one cannot infer that
John is a heathen (one can infer that he was one).
These two characterizations of how grammatical form is misleading
as to logical form have the virtue of bringing out some further aspects
of the notion and doctrine of logical form. For instance, they can be
used to explicate, and motivate the study of, ideal languages and can be
shown to explicate a wide variety of classical examples of ways in
which grammatical form is misleading as to logical form (see Hamish
1972, ch. 1). If this is a sound approach, traditional wisdom on logical
form need not rule out a purely inferential account of semantics. The
notion of the misleadingness of grammatical form can be explicated
without invoking forms to compare. To speak of the logical form of a
sentence is, on this view, just an odd (and misleading) way of speaking
of a particular kind of description of a sentence (like describing a sen-
tence as being of the form NP + VP); it is not another object to be
described.
Thus, the possibility remains that inference rules are a main, perhaps
the main, device in a semantic theory after theoretical definition. Infer-
ence rules would be a part of the apparatus used by a hearer in the
interpretation of a speaker's remarks. What the rules are and how
notions like synonymy and analyticity might be defined in terms of
them are matters of future research. This conclusion still leaves at least
one question unanswered: with no lexical decomposition (beyond defi-
nitional abbreviation) what does meaning specification amount to?
Fodor (1975) has argued that the vocabulary of the language of thought
is roughly equivalent (in size) to that of a natural language, and that it is
the system of unambiguous, psychologically real semantic representa-
tion. This hypothesis is compatible with the absence of decomposition,
but doesn't the hypothesis still require a distinct system of semantic
representation? Fodor (1975) argues that it does, but perhaps some-
thing weaker will do (see Hamish 1977b). It could be that many sen-
tences in the language of thought are regimented forms of the language
149
Philosophical and Linguistic Issues
one speaks, with entailment (or translation) principles connecting them
with nonlinguistic modes of representation. This hypothesis would
allow for the possibility of learning a language in which one eventually
thinks and would not necessitate a system of semantic representation
independent of the vocabulary (and syntax) already required in a
grammar. Of course we have not explained the nature of such rep-
resentation, linguistic or nonlinguistic, but explaining it can be consid-
ered a problem for future research (see Rosenberg 1974). In any event
its outcome does not effect the SAS.
Finally, theories of semantic representation usually contain, im-
plicitly or explicitly, definitions of the semantic properties and relations
predicted to apply to the expressions they cover. Thus, synonymy (on a
reading) might be defined as sharing a semantic representation. How
would such a theory define meaningful (as opposed to meaningless)
expressions? A typical definition would run as follows:
(11) Expression e has a meaning (is meaningful) if and only if grammar
G assigns e some semantic representation.
This will not do as a general definition of being meaningful, for the
obvious reason that the grammar G might be wrong about e. If gram-
mar G failed to assign "It is raining" a semantic representation, it
would not follow that the sentence is not meaningful (is meaningless)
but only that grammar G is wrong. 7 And when a grammar assigns a
semantic representation to an expression, thereby indicating that the
expression means something (whatever the representation represents),
it does not say what this pairing amounts to. Formally, the pairing of
sound and meaning in a grammar could record all sorts of things. Thus,
definition (11) cannot be used as evidence for the existence of a level of
semantic representation.
We saw in section 2.2 that L1(a) in the SAS requires representing
that e means ... (as well as perhaps __ ) in L. In this section we have
been looking at issues surrounding the problem of what kind of seman-
tic representation should replace the ellipses. We have given some
reasons for preferring an inferential approach to a decompositional
one, quite apart from its being more consonant with the SAS.
8.2. SPEAKER MEANING
So far, we have tried to avoid using the term speaker meaning (as
opposed to "expression meaning" or "sentence meaning") for two im-
portant reasons. First, a number of distinct things go by the label
Issues 150
"speaker meaning" and we have not yet sorted these out. That will be
our first task here. Second, as we will show, there is no reason to
suppose there to be a single, stable, pretheoretic notion of speaker
meaning capable of bearing the theoretical weight put upon it in many
analyses.
8.2.1. Varieties of Speaker Meaning
At least five different things can be meant when it is said that somebody
means something. (1) When we say that somebody meant what he said,
we may mean that he is serious or sincere about it. Or (2) we may mean
that he is speaking literally: "S said that p and meant it" is ambiguous
between the seriousness and the literalness of S's utterance, depending
on whether emphasis is on "meant." Generally, when a person is
speaking literally, he is also speaking seriously, but it is possible to be
literal without being serious ("I was just kidding"), just as it is possible
to be serious without being literal ("Not until Hell freezes over").
However, we probably would not say that S meant what he said if he
was speaking seriously but not literally.
(3) In a third use, "means" has the sense 'intending': in saying
something a person may mean (intend) to be doing such-and-such. This
may be an illocutionary act, or a perlocutionary act, or even some
collateral act. 8 (4) Then there is operative meaning, as used in line L2 of
the SAS, which specifies how a speaker is using an expression (word,
phrase, or sentence). Such a specification is of the form: S meant ... by
e. This use is to be contrasted with (5) the notion of speaker meaning
proper, as given by: In uttering e, S meant thatp-which Grice (1957,
1969) introduced. 9
It is important to distinguish these five uses of "means" lest we fall
into hopeless confusion in trying to understand what it is for a person to
mean something by an utterance ("utterance" itself suffers from act!
object and type/token ambiguity). Confusing uses (4) and (5) collapses
an essential distinction in the theory of speaker meaning. Frye, in an
important article (1973) concerned with such distinctions, herself fails
to distinguish (4) and (5), or perhaps singles out (4) while ignoring (5).
She is explicit about use (3), and points out that Searle confuses (3) and
(4) when he says of his essential condition on promising that it "cap-
tures our amended Gricean analysis of what it is for the speaker to
mean to make a promise" (1965, 237). Searle's revised Gricean analysis
Philosophical and Linguistic Issues 151
(1969, 49-50) is explicitly of "the different concept [from Grice's
meaningnn] of saying something and meaning it," which Searle glosses
as our use (2), meaning literally what one says. to
It might seem that uses (3) and (5) are coextensive: that to mean (3) to
be performing some illocutionary act F(P) is to mean (5) that p. For
example, it might be thought that-except for illocutionary acts that
lack propositional content (utterance of "Hello"), which are hardly
candidates for meaning (5) any thing- "means" is equivalent to "in-
tends to be performing some illocutionary act," that is, that "means"
(5) falls under "means" (3). The trouble is that there seems to be no
way of filling in the blank in: S means that p-that is, in specifying p
for certain sorts of illocutionary acts with propositional content. There
is no problem with constatives and directives (these correspond to
Grice's (1969) indicative and imperative cases). However, it is not clear
what S means when issuing an acknowledgment, even one that ostensi-
bly has propositional content like "Congratulations on getting pro-
moted" or "Thanks for the wonderful time." One way out, for which
there are independent reasons, is to deny that these really have prop-
ositional content. But still there is the case of commissives. Is there
something of the form, "that p," which S means (5) when he promises
to return? If S meant merely that he would return, his utterance would
not be distinguished, as far as meaning (5) is concerned, from making a
prediction or from merely expressing an intention that he would return.
On the other hand, it is not clear that S meant (5) that he promised that
he would return, even if he meant (1), (2), or (3) that he promised.
8.2.2. A Critique of Speaker Meaning
Let us now examine the notion of speaker meaning proper, that is,
meaning (5). Illocutionary acts are performed either with R-intentions
or with intentions implemented by illocutionary conventions, C-inten-
tions. Clearly those acts with only C-intentions are not cases of speaker
meaning, since speaker meaning requires R-intentions not C-intentions.
Conventional illocutionary acts have as their primary illocutionary
point the changing or as-if changing of institutional states of affairs, and
insofar as they involve uptake, uptake is more verificatory than con-
stitutive of success (see section 6.3.2). Even restricting speaker mean-
ing to communicative illocutionary acts, for reasons mentioned at the
end of section 8.2.1, we should further restrict the present discussion to
 Issues 152
constatives and directives, corresponding to Grice's indicative and im-
perative acts of meaning, respectively, and to Schiffer's cases of
"meaning that p" and of "meaning that A is to '1'."
The most glaring question about the notion of speaker meaning is
whether there really is a sense of the term "meaning" that corresponds
to just the idea of "attempted communication in the sense which Grice
seeks to elucidate" (Strawson 1964a, 447). It is possible, one starts
surmising after surveying the literature, that the endless parade of
counterexamples to successive analyses designed to accommodate pre-
vious counterexamples betokens a spurious notion. This sentiment is
intensified when one examines details of successively more refined
analyses: more and more subtle intentions are ascribed to speakers,
and the ability to recognize such subtler and subtler intentions is as-
cribed to hearers, not to mention S's ability to ascribe this ability to H
and H's ability to ascribe the attribution of this ability to S, and on and
on. Intentions are proliferated ad infinitum, but allegedly harmlessly.
Alternatively, what is required is mutual knowledge, with an allegedly
harmless infinite regress condition, of a single, but awesomely com-
plex, intention (given by Schiffer 1972, 63). One wonders, simply, just
what these proposed analyses are analyses of.
The difficulties become evident when one considers Schiffer's objec-
tions to, first, the alleged sUfficiency and, second, the alleged necessity
of Grice's analysans. Objections of the first sort give rise to the addition
of more and more intentions to the analysans (later replaced by Schif-
fer's mutual knowledge condition). The second sort pertain to the ne-
cessity of Grice's "by means of recognition of intention" condition and
to the specification of the intended response by the hearer. Roughly
speaking, considerations of the first sort raise the question of how much
it takes to mean something; considerations of the second sort concern
what one is doing when meaning something.
Schiffer's objections to the alleged sUfficiency of Grice's analysans in-
volve raising a counterexample to the analysans and to successive modi-
fications incorporating further intentions inspired by previous counter-
examples. And, says Schiffer (1972, 26), "What makes each of the
examples ... a counter-example is that S intends to deceive [H] in one
way or another," namely, as to one or another of S's intentions. The
general pattern of successive modification is, indeed, that each addi-
tional intention be that H recognize some previous intention of S' s.
Before arriving at his replacement of the iterated intentions with the
Philosophical and Linguistic Issues 153
mutual knowledge condition, Schiffer notes Grice's efforts to deal with
the regress either by a condition that instead of adding more and more
intentions simply requires that no deceptive intentions be present or by
claiming that there is a de facto limit on the number of intentions that
can realistically appear on the list: a limit to the subtlety of human
intellect. Although Schiffer's mutual knowledge condition is a substan-
tial improvement on Grice's iterated intention conditions, nevertheless
its point is still to preclude "counter-examples based on deception . . .
what.precludes these cases from being instances of [speaker] meaning
is that S does not utter e expecting that if the intentions with which he
uttered e are satisfied, [H] will recognize that S meant something by
uttering e" (Schiffer 1972, 41).
A curious fact about these various attempts to define speaker mean-
ing by ruling out more and more subtle cases of deception is that simple
deception is not precluded at all. By simple deception we have in mind
cases where S means that p but doesn't believe that p and where S
means that H is to '1' but doesn't want H to '1'. Simple deception, or
insincerity, does not involve S's having any intentions that H is not to
recognize, only that S have a certain belief or desire that H is not to
recognize. On the one hand, it seems clear that any analysans permit-
ting S to deceive (simply) cannot be an analysans of meaning, at least
not if meaning implies sincerity. However, it might be argued that
meaning permits insincerity - we are not talking about meaning in the
sense of "saying something and meaning it," as Searle (1969, 49), for
example, seems to think-but meaning in the sense of communicating,
irrespective of whether one believes what one is communicating. But if
simple deception is allowed, why should meaning preclude any of the
more subtle forms of deception that involve hidden intentions? Be-
sides, if these cases, as given in the various counterexamples, aren't
cases of meaning, what are they cases of? It seems that each successive
refinement of the analysans of meaning widens the gap between clear
cases of nonmeaning, ruled out by Grice's original analysis, and clear
cases of meaning, as provided by the analysans in question. What,
again, are we to call these intermediate cases?
Schiffer's objections to the alleged necessity of Grice's analysans
concern two things, first, Grice's "by means of recognition of inten-
tion" requirement, and second, the "production of belief or action
(intention)" requirement. The first need not concern us here, for it
simply points out the excessive narrowness of Grice's requirement,
Issues 154
which Schiffer broadens to a "by means of recognition of connection"
requirement, as we might call it, between the utterance and the in-
tended response. Only in some cases of speaker meaning does the
relation that H is to recognize between e and the intended response
have to be the relation of being intended by S to be recognized by H as
related.
However, the production of belief or action requirement is much too
strong, and hardly necessary, even if qualified by the activated belief
or intention stipulation. The reason is simple: for S to succeed in mean-
ing something, H must understand what S means and nothing more.
Understanding what S means does not require, if S means thatp, for H
to believe that p, or if S means that H is to 'l', for H to 'l' or intend to.
Therefore, this requirement is too strong. Searle is quite right when he
charges Grice with confiating illocutionary and perlocutionary effects
(1969, 46), and the charge applies to Schiffer as well. An adequate
account of meaning and speech acts must distinguish the two types of
effects, even if, in general, no illocutionary act is performed without an
intended perlocutionary effect.ll
Finally, if Schiffer's (1972, 63) analysis were correct, his mutual
knowledge condition would be incompatible with the requirement of
apparent sincerity. If the analysis implies that S is sincere and also
requires that S's primary intention be recognized on the basis of mutu-
ally known conclusive evidence (and mutually known to be conclusive
evidence), it would seem to follow that S cannot mean something while
being insincere, even if his insincerity is not recognized and H believes
S is sincere. For if S is insincere (and, presumably, knows it), he cannot
consistently believe, as required by the mutual knowledge condition,
that the obtainment ofE, as realized by his utterance of e, is conclusive
evidence that he uttered e with the primary intention that there be some
reason for which H responds (as S's utterance of e is supposed to cause
H to respond). For from S' s point of view , there can be no such reason.
Hence S cannot even have that primary intention. Instead, his primary
intention is, inter alia, to provide H with a bad reason for his response.
And surely the badness of this reason cannot be mutually known if it is
still to serve as a reason for H's response. What seems to be the
contradiction immersed in Schiffer's analysis is that meaning does re-
quire sincerity after all, even though it should not, since the sense of
meaning that implies sincerity is not the one in question, speaker
meaning.
Philosophical and Linguistic Issues 155
8.3. PRESUPPOSITION
In this section we examine the variously drawn contrast between what
is said and what is presupposecf:, We suggest that presuppositional facts
can be handled in terms of independently motivated notions from the
SAS.
In the past decade the notion of presupposition has been applied to a
multifarious collection of phenomena: necessary conditions for truth-
valuation (Lakoff 1972, V), felicity conditions of speech acts (Fillmore
1971), shared information (Jackendoff 1972), and conditions necessary
for an utterance to be meaningful (Muraki 1972). Though widely uti-
lized, presupposition has been rarely scrutinized in linguistics (see
Kempson 1975; Wilson 1975; Katz and Langendoen 1976). One might
well suspect its popularity is due primarily to its adaptability. Such
wanton adaptability can rob the notion of most of its predictive and
explanatory value,1 2 It is not our ambition to untangle the various
conceptions (and misconceptions) of presupposition. Rather, we want
to locate the relevant phenomena in our theory and see whether they
pose any problem for the SAS, as regards either what it covers or what
it omits. To this end we will discuss the two main species of presuppo-
sition (as we understand it), each of which contrasts with what is
said, entailed, and implied. These are semantic and pragmatic pre-
supposition.
8.3.1. Semantic Presupposition
The central feature of semantic presupposition, as introduced by Frege
(1892)13 and revived by Straws on (1950, 1952),14 is that the presupposi-
tions of a statement made in uttering a sentence (the SMU of a sen-
tence) are referential conditions that must be satisfied for the SMU of
the sentence to be either true or false. Thus, Frege contended that the
SMU of the sentence (13a) does not include but presupposes (13b)1S-
(13) a. Kepler died in misery.
b. Kepler exists.
-because he also held that (14),
(14) Kepler did not die in misery.
is used to make the contradictory statement of (13a), and that the SMU
of (14) presupposes that Kepler exists (13b) , and so bears the same
Issues 156
relation to (13b) as (13a) does. In other words, the SMU of a sentence
and the SMU of its contradictory have the same presupposition.
Straws on' s extension of the presuppositional data beyond Frege's
was fairly conservative. Like Frege he includes sentences with singular
definite descriptions, and so the SMU of (15a) is said to presuppose that
(15b):
(15) a. The present King of France is bald.
b. There is only one present King of France.
But Strawson (1952, 173-179) also counts quantified plural referring
expressions as carrying presuppositions,16 so that the SMU of (16a)
presupposes that (16b):
(16) a. All John's children are asleep.
b. John has children.
Thus the original core conception of presupposition had the follow-
ing characteristics:
Presupposition:
(a) The presuppositions of the SMU of a sentence s concern conditions
of reference on the truth-valuation of the SMU of s.
(b) The SMU of a sentence s and its contradictory not-s have the same
presuppositions. 17
If we assume that certain conditions are satisfied if a statement ex-
pressing such satisfaction is true, then we can add a third character-
istic:
(c) The truth of the presupposition of SMU of s is guaranteed (ne-
cessitated) by:
i. the truth of the SMU of s, and
ii. the falsity of the SMU of s.
iii. If its presuppositions are not true, the SMU of s has no truth value.
The classical conception of (statement) presupposition held that the
SMU of sentences such as (13a)-(15a) presuppose (13b)-(15b) re-
spectively; and that the relation of presupposition has characteristics
(a)-(c).
As thus characterized, how would semantic presupposition be rep-
resented in the speech act schema? Should presupposition be included
in the meaning specification of a sentence whose SMU has a presup-
position? One could bifurcate semantic representations into two parts,
Philosophical and Linguistic Issues 157
one specifying what is being presupposed and one specifying the re-
mainder (see Katz 1977c, ch. 3). However, (operative) meaning con-
tributes to what is said (see chapter 2), so if presupposition is a part of
sentence meaning but not a part of what is said, the schema must be
supplemented with a procedure for selecting only nonpresuppositional
aspects of meaning as contributing to what is said. Since this complica-
tion of the schema would not be necessary if there were no semantic
presupposition, it is worth inquiring whether there really is such a
phenomenon as semantic presupposition. We think not, though here we
cannot fully document our reasons.
First, considered purely as linguistic judgments, the data are not
really as clear as usually thought. Fluent speakers show considerable
variation when asked to judge whether the SMU of a sentence is false
or neither true nor false, when the putative presupposition fails. And
even if tests were devised which showed stable responses to sentences
like (13a)-(15a), there seem to be clearly related sentences whose
statements speakers judge not to have presuppositions of the relevant
sort. For instance, the SMU of (17a) does not presuppose that (17b)-
(17) a. Pegasus was ridden by Gene Autry.
b. Pegasus exists.
-because knowing that (17b) is false, speakers judge that (17a) is
simply false. The same goes for (18b) and (18a) and for (19b) and (19a):
(18) a. The present King of France sold you a vacuum cleaner.
b. There is just one present King of France.
(19) a. All of John's children came for dinner.
b. John has children.
Why is it the case that the SMUs of (17a)-(19a) are judged as non-
presuppositional? We do not know for sure, but notice that these sen-
tences are all overtly relational, and the verbs are all transitive,
whereas the original (13a) -(16a) are purely predicative (in surface
form). Perhaps when a putative object is claimed to bear some relation
to something else, it counts toward the falsity of that claim that the
object does not exist.
Even if there were stable judgments about such sentences and a
presuppositional theory could account for them, we could still ask
whether a presuppositional theory is the best account of these judg-
ments. Deciding which is the best theory involves settling questions of
overall explanatory power and simplicity. Since such a theory would
Issues 158
have to account for the full range of relevant data, it would have to
deal with the presuppositions of compound sentences as well, as in
Karttunen (1973). However, recent work by Kempson (1975), Wilson
(1975), and Boer and Lycan (1976) suggest that what has been called
"semantic presupposition" is better viewed as a special case of entail-
ment, plus pragmatic presupposition (but see Katz 1977c). Entailment
is a semantic relation already utilized in our discussion of the locution-
ary act portion of the schema. Thus, if the notion of pragmatic presup-
position relevant for the explanation of semantic presupposition can be
explicated in terms of concepts already developed in the schema, the
phenomenon of semantic presupposition will be accounted for with-
out complicating the basic structure and conceptual resources of the
schema. What, then, is the relevant notion of pragmatic presupposition
that can, in conjunction with entailment, help account for the facts
commonly thought to be cases of semantic presupposition?
8.3.2. Pragmatic Presupposition
There appear to be three main kinds of pragmatic phenomena labeled
"presupposition" in the literature. For neutrality we label them with
numbers. In each case the (b) sentence is thought to be a pragmatic
presupposition of the (a) sentence.
(Pragmatic) Presupposition1
One conception of presupposition is that it concerns speaker's assump-
tions (beliefs) about the speech context. As Lakoff writes, "Natural
language is used for communication in a context, and every time a
speaker uses a sentence of his language . . . he is making certain as-
sumptions about that context" (1970, 175). As examples of such phe-
nomena we find factives and aspectuals:
(20) a. Sam realizes that Irv is a Martian.
b. Irv is a Martian.
(21) a. Sam does not realize that Irv is a Martian.
b. Irv is a Martian.
(22) a. Sam has stopped beating his wife.
b. Sam was beating his wife.
(23) a. Sam has not stopped beating his wife.
b. Sam was beating his wife.
Philosophical and Linguistic Issues 159
(Pragmatic) Presupposition2
Another notion of (pragmatic) presupposition is that the (pragmatic)
presuppositions of a sentence are those conditions that have to be
satisfied in order for the intended speech act to be felicitous and appro-
priate in the circumstances. Keenan writes:
In general I want to consider that the presuppositions of a sentence are
those conditions that the world must meet in order for the sentence to
make literal sense . . . Now how many sentences require that certain
culturally defined conditions or contexts be satisfied in order for an
utterance of a sentence to be understood . . . these conditions are nat-
urally called presuppositions of the sentence . . . An utterance of a
sentence pragmatically presupposes that its context is appropriate.
(1971,45,49)
Fillmore puts the matter another way:
By the presuppositional aspects of a speech communication situation, I
mean those conditions which must be satisfied in order for a particular
illocutionary act to be effectively performed in saying particular sen-
tences. (1971, 276)
As examples of such phenomena we are given the following:
(24) a. John accused Harry of writing the letter.
b. There was something blameworthy about writing the letter.
(Fillmore 1971)
(25) a. John criticized Harry for writing the letter.
b. Harry wrote the letter. (Fillmore 1971)
(26) a. Tu es degoutant.
b. The addressee is an animal or child, socially inferior to the
speaker, or personally intimate with the speaker. (Keenan 1971)18
(Pragmatic) Presupposition3
A third notion of (pragmatic) presupposition is that of shared (or back-
ground) information: "We will use ... 'presupposition of a sentence'
to denote the information in the sentence that is assumed by the
speaker to be shared by him and the hearer" (Jackendoff 1972,230). As
Bates puts it, "Presupposing is the act of using a sentence to make a
comment about some information assumed to be shared or verifiable by
speaker and listener" (1976, 25). As examples of such phenomena we
are given:
Issues 160
Philosophical and Linguistic Issues 161

(27) a. Was it Margaret that Paul married? (31) a. John realizes that his car has been stolen.
b. Paul married someone. b. John doesn't realize that his car has been stolen.
(28) a. Betty remembered to take her medicine. c. John's car has been stolen.
b. Betty was supposed to take her medicine. (32) a. John has stopped playing tennis.
(29) a. That Sioux Indian he befriended represented the Chief. b. John hasn't stopped playing tennis.
b. He had befriended a Sioux Indian. c. John was playing tennis.
(30) a. He befriended that Sioux Indian who represented the Chief.
b. Some Sioux Indian represented the Chief. We account for the necessitation of (c) by (a) in terms of bidirectional
entailment:
These three notions of pragmatic presupposition are loosely related.
Presuppositionl and presupposition3 overlap in that if p is assumed ~y (33) i. x realizes that p iff:
the speaker to be shared, it must be assumed by the speaker and so be a ii. x believes that p & p.
case of presuppositionl. Likewise, if a certain condition is necessary (34) i. x has stopped cf>-ing iff:
for the successful and felicitous performance of an illocutionary act, ii. x was cf>-ing & x is not now cf>-ing.
then in general that condition must be believed by the speaker to obtain
And if the presumption of sincerity is observed, the speaker implies
and so be a special case of presuppositionl. The main problem with
that he believes that p, and that x was 'l'-ing.
presuppositionl is that it is too inflationary with respect to contexts. If
But how are we to account for the presupposition on the negative (b)
someone presupposeSl that Irv is a Martian (that is, from the planet
sentences? Entailment plus the presumption of sincerity will not do
Mars), then the belief that this is true is a belief about context-so
because the negative sentences do not entail the presupposed (c) sen-
Mars is a part of the context. If, on the other hand, one adopts an
tences, for the semantics of the negatives are disjunctive:
entailment analysis of factives and aspectuals, then presuppositionl (in
the positive case) is an entailment and belief in the presuppositionl is a (35) i. x does not realize that p iff:
consequence of the presumption of sincerity. This leaves presupposi- ii. ~x believes that p v ~p.
tion2 and presupposition3 to account for. (36) i. x has not stopped cf>-ing iff:
Presupposition2 seems to amount to success and felicity conditions ii. ~x was cf>-ing v x is now cf>-ing.
on speech acts, no more, no less (see section 2.5). As such it is a part of
The clue to our answer is noticing a similarity between the semantics of
the general theory of speech acts and does not require any modification
the positive sentences and the way one normally understands their
of our theory. How about presupposition3? Pretty clearly for us, back-
(internal) negations. The negatives give the understanding: 20
ground assumptions made by S are simply assumptions made by S as to
what is currently a mutual belief.19 By accounting for the data sup- (35') i. x does not realize that p:
porting the notion of semantic presupposition in terms of entailment ii. ~x believes that p & p.
together with these conceptions of pragmatic presupposition, we need (36') i. x has not stopped cf>-ing.
not modify the SAS to handle presupposition: we can do without ii. x was cf>-ing & x is now cf>-ing.
semantic presupposition since the phenomena of pragmatic presup-
position are handled by independently motivated (and independently Suppose that p or that x was cf>-ing is assumed by the speaker to be
labeled) aspects of the schema. For this reason we avoid the term mutually believed. Borrowing a formal device from Grice (1967, ch. 4),
we can bracket off these clauses:
pragmatic presupposition except when discussing the views of others.
In summary, sentences like (a) and (b) in (31) and (32) have been said (37) x realizes that p: x believes that p [& pl.
to presuppose the (c) sentences, where this has been explicated either (38) x has stopped cf>-ing: [x was cf>-ing &] x is not now cf>-ing.
as necessitation of or as belief in their truth.
Issues 162
Given that bracketed material is assigned the status of a mutual belief
it will be resistant to negation-the brackets tend to restrict the scop;
of such an operator. The result of negating (37) and (38) without pene-
trating the brackets (and thus violating its common ground status) is the
following:
(39) x does not realize that p: ~x believes that p [& p].
(40) x has not stopped cp-ing: [x was cp-ing &] x is now cp-ing.
Notice that these are just the normal understandings we earlier re-
corded in (35') and (36'). We can now accommodate two facts. First,
the negative factives and aspectuals are understood in such a way that
the semantic presuppositions are accepted as true by the speaker. Sec-
ond, one can always explicitly deny the bracketed material, and still
speak truly:
(41) x does not realize that p because ~p.
(42) x has not stopped cp-ing because x never started cp_ing. 21
These can be true in virtue of the disjunctive semantics of the clause
before "because":
(41') (~x believes thatp v ~p), because ~p.
(42') (~x was cp-ing v x is now cp-ing), because ~x was cp-ing.
The bracketing device is encouraging, but still there are some ques-
tions to be answered about it. For instance, how does material get
assigned to or marked as common ground? We propose that it comes in
two steps. First, expressions have a certain bracketing for internal
linguistic reasons or by historical accident. Then, there is a rule 22 of
language use to the effect that,
Bracketing (Br): S is to utter e containing bracketed [q] in saying that
*( . ..p ... ) just in case S believes that q is mutually believed by Sand H.
S is not to utter a sentence with a factive or aspectual predicate unless
he takes it to be common ground that [q], other things being equal. Of
course context (linguistic and nonlinguistic) can force negation into the
bracketed material, and so brackets are defeasible. This is the result we
want.
How does an element get bracketed in the first place? Sometimes
there are semantic reasons. In (43) both (a) and (b) presuppose (c):
Philosophical and Linguistic Issues 163
(43) a. John has finished practicing.
b. John has not finished practicing.
c. John started practicing.
If the semantics of (43a) and (43b) are
(43') a. x started cp-ing & x completed cp-ing,
b. ~x started cp-ing v ~x completed cp-ing,
then the bracketing of (43a) has to be as in (44) because the second
clause entails the first.
(44) [x started cp-ing &] x completed cp-ing.
This rules out the other possible bracketing, represented in (45),
(45) x started cp-ing [& x completed cp-ing]
because the result of negating (45) and not penetrating the brackets
would result in the self-contradictory (46): 23
(46) ~x started cp-ing [& x completed cp-ing].
Thus there is internal semantic reason for the bracketing to be as we
have postulated it. With "stop," however, there is no such semantic
dependency in its analysis, so this cannot be the explanation of why its
bracketing is indicated in (40). Notice, though, that the element as-
signed common ground status is time-indexed for the past, and that
does make it more suitable as something already believed (assumed to
be mutually believed). But with "realize" neither of these expression-
internal reasons can be used to account for its bracketing-there is no
entailment nor temporal asymmetry between conjuncts. Perhaps it is
an historical accident that the truth clause and not the belief clause is
bracketed, though we suspect not. The same bracketing pattern occurs
in a variety of other cases, as we will see.
Finally, why should there be such a device as bracketing at all? What
might its status and role in communication be? Assume that many
con~ersations are governed by the presumptions of not being overly
a~blguouS and nonspecific. Negative factives, on the standard analy-
SIS, have the form of disjoined negations. If the speaker meant to com-
m~nicate such a disjunction, he would be flouting the presumption, for
~Ithout bracketing the denial would still be unspecific, ambiguous, or
mdeterminate. The bracketing device (in the negative cases) reduces
Issues 164
ambiguity and increases specificity, since the negative operator goes to
just one condition.
What is the status of this bracketing device? Is it a part of the lan-
guage, a conventional device that serves a purpose like the one
sketched above? Or is it just a principle of interpretation based on
conversation or discourse? To us the latter seems more plausible; the
bracketing reflects normal expectations about what a speaker intends
to communicate. These expectations are based mainly on previous
experience with the point or topic of such remarks in the past. If these
experiences were to change, expectations would change and so in
many cases the bracketing would change. This account has the virtue
of meshing in an obvious way with our earlier justification for the
device, and it also meshes with the fact that in certain circumstances,
context can force a change in bracketing. In a context in which it is
obvious that x believes that p (xBp), the negative factive will not be
interpreted as the denial of xBp. The general principle seems to be:
Operator Scope (0): If C is a condition in the analysis of an expression
e, and C is contextually satisfied, then the operator is taken as going to
the next most deeply embedded condition consistent with what is
meant and with the context.
Thus, if the context is such that it is clear that xBp, then the principle
(0) predicts that the negative factive will be taken as a denial that p.
With some reservations we suggest that this proposal can be ex-
tended in three directions: to other operators besides negation, to other
factives besides "realize," and to nonfactives (see Hamish 1976b,
374-376, for discussion of complexities of individual cases).
Other Operators: It seems that, at least in an unbiased context, sen
tences containing the following operators would be taken primarily as
remarks about belief:
possibly: Possibly x realizes that p. (possibly xBp ... )
unlikely: It's unlikely that x realizes that p. (unlikely that xBp ... )
uncertain: It's uncertain that x realizes that p. (uncertain that xBp ... )
must: x must realize that p. (x must Bp ... )
finally: Finally x realizes that p. (finally xBp ... )
Other Factives: The same sort of analysis and bracketing seems to
work with other factives like "recognize," "remember," "be aware,"
"admit," and "know." For instance, if we suppose the following to be
roughly correct,
Philosophical and Linguistic Issues 165
(47) x is aware that p iff xBp [& p]
then the denial of (47) has just the force we take it to have-as the
denial that x believes that p. The same seems to hold for one common
propositional use of "recognize." And if we suppose "admit" to have
an analysis something like
(48) x admits that p iff x states that p [& p],
then the negative factive, "x didn't admit that p," comes out right, as
primarily a denial that x stated (or would state) that p. Perhaps other
cases could be handed as these are.
Other Nonfactives: As a final application we reanalyze some of
Fillmore's verbs of judging (1971, 188-189), though we do not claim
(nor does Fillmore) that these analyses are adequate as they stand:
(49) x accused y of cf>-ing iff x stated that y cf>-ed [& x believes y's cf>-ing
is blameworthy].
(50) x criticized y for cf>-ing iff x stated that cf>-ing is blameworthy [& x
believes y cfJ-ed].
(51) x blamed y for cf>-ing iff x stated that y cf>-ed [& x believes y's cf>-ing
is blameworthy].
This is only a sample of the possible range of phenomena susceptible to
a bracketing analysis. 24 A classical case like definite descriptions is a
further example and can be handled similarly.
We conclude from this glimpse at presupposition, both semantic and
pragmatic, that the SAS does not need to be elaborated just to accom-
modate presuppositional data. So-called presuppositional facts can and
should be analyzed in terms of independently motivated notions from
the schema.
8.4. IMPLICATURE
We need to contrast what is said not only with what is presupposed but
also with what is implied. In an important series of papers, published
and unpublished, Grice has been developing a theory of the relation-
ships among an expression, its meaning, the speaker's meaning, and the
implications of the utterance. The relevant categories are indicated by
the tree in figure 8.2.25 One convenient way of investigating Grice's
theory is by tracking down the implicature branches of the tree.
In a number of places (1961, 444; 1967; 1968, 225), Grice has at-
tempted to draw, both pretheoretically and theoretically, a distinction
Issues 166 Philosophical and Linguistic Issues 167

Total content (signification) of an utterance (remark) Cooperative Principle (CooP): "Make your conversational contribution
such as is required, at the stage at which it occurs, by the accepted
purpose or direction of the talk-exchange in which you are engaged"
what was meant what was said what is implicated (Grice 1975, 45).
Under this principle come the maxims 29 of quantity, quality, relation
(relevance),30 and manner. The first three pertain to what is said, the
what is conventionally what is nonconventionally what was fourth to how what is said is said. 31 These maxims are essentially im-
implicated implicated presupposed peratival versions of the conversational presumptions of quantity,
quality, relevance, and manner respectively, which we have formu-

/"""
what is conversationally
implicated
"'- "'-
"-
"'-
"'- "-
"'-
particularized generalized conversational
conversational implicatures implicatures
Figure 8.2 Relationships among an expression, its meaning, the speaker's
meaning, and the implications of an utterance
between what someone stated or said on an occasion, and what was
implied or implicated. 26 To our knowledge, he has not attempted to
formulate necessary andjointly sufficient conditions for implicating nor
to give a general procedure for identifying, on an occasion, what has
been implicated. 27 Rather, he has been concerned mainly with sorting
out different kinds of implicature and analyzing their modes of opera-
tion. Grice distinguishes three major kinds of implicature: conventional
implicature, conversational implicature, and presupposition. He has
little to say about the first type, and we have taken up the last category
already; we will therefore look only at his views on the second type,
conversational implicature.
In contrast with conventional implicatures (which turn on the mean-
ings of the words used) there is a class of implicatures that turn not only
on what a person says but also on principles governing disc()urse.
Grice's theory is the latest, and most sophisticated, in a line of attempts
to account for what has been called contextual or pragmatic implication
(for a survey of earlier efforts, see Hungerland 1960). Grice's account
applies to discourse governed by the cooperative principle: 28
lated in section 4. 1.
According to Grice, there are many different ways for participants in
a talk-exchange to fail to observe a maxim. Since Grice has no one term
for this, we adopt the neutral term infringement for any failure to fulfill
a maxim (or the CooP). Saying that A infringed a certain maxim means
simply that he failed to fulfill it. This use commits us to nothing regard-
ing the way the maxim was not fulfilled or the consequences of not
fulfilling it. Of the four ways Grice discusses, only three of them give
rise to implicatures. 32
First, one may opt out by indicating plainly that he is unwilling to
cooperate ("I cannot say anything more"). This seems to be the only
infringement that does not give rise to implicature.
Second, one may quietly and unostentatiously violate a maxim. In
violating a maxim one is likely to mislead. Grice does not say very
much about this as a way of infringing a maxim distinct from the next
two. In fact, sometimes he uses violate in the general way we have
reserved for infringe. Opting out is logically distinct from violating.
That is, if S opts out of a maxim, he does not infringe it "quietly and
unostentatiously" and thereby violate it. Conversely, if S violates a
maxim, he has not indicated plainly that he is "unwilling to cooperate"
and so is opting out.
Third, one may be faced with a clash between one maxim and an-
other. In this case a maxim may be infringed, but its infringement is
explained by supposing it to conflict with another maxim. Consider
Grice's example. Suppose A is planning an itinerary with B for a holi-
day in France. Both know that A wants to visit his friend C.
(52) A: Where does Clive?
B: Somewhere in the south of France.
Gloss: (a) B is not opting out . (b) B has infringed the maxim of
Issues 168
quantity (say as much as necessary). (c) Observation (b) can be
explained only by supposing that B is aware that to be more infor-
mative would be to infringe the maxim of quality (have evidence
for what you say). (d) So B implicates that he does not know which
town C lives in.
lt is in the nature of the clash that under the circumstances the speaker
cannot fulfill both of the maxims in question at once. However, Grice
does not claim that any particular maxim must override any other. In
this case, the maxim of quality overrides the maxim of quantity. Notice
also that Grice claims that the supposition in (c) is the only one that
could explain (b), but he does not spell out why.
Let us view this clash from both speaker's and hearer's standpoints.
From the speaker's point of view, under the circumstances S must
infringe either the maxim of quantity or the maxim of quality and so he
must make a choice between giving not enough information and giving
groundless information. If the speaker does not opt out of the maxim of
quality (by saying "I do not know exactly"), the hearer is faced with an
infringement he can explain (or explain away) by positing a clash and
supposing the speaker opted for fulfilling the maxim of quality. This
suggests that there is an ordering or weighting of the maxims that can
be explained as follows. If it can be assumed that the speaker S is
observing at least CooP, then S will pick quality over quantity if only
because truly groundless information has at least as good a chance of
being wrong as right, and as such would probably not be helpful,
thereby violating CooP.
On the hearer's side, there seems to be a corresponding metaprinci-
pIe at work, the principle of charity:
Principle of Charity (PC): Other things being equal, construe the
speaker's remark so as to violate as few maxims as possible.
However, since some maxims may be more highly weighted than
others, we need a weighted principle of charity (for pairs of maxims):
Weighted Principle of Charity (WPC): Other things being equal, con-
strue the speaker's remark so that it is consistent with the maxim of
higher weight. Or, if the speaker has infringed one or other of a pair of
maxims, other things being equal, assume that he has infringed the
lowest valued maxim.
Using WPC it is possible to reconstruct Grice's gloss in more detail
and to tighten the appeal to explanations (see Harnish 1976c, 344).
Philosophical and Linguistic Issues 169
Fourth, the speaker may flout a maxim-he may blatantly fail to
fulfill it. S exploits a maxim when he flouts it, with the consequence
that H must reconcile S's saying what he said with the supposition that
S is obeying the conversational maxims and the cooperative principle.
With this in mind we can now characterize conversational implicature
more precisely.
Conversational Implicature: S conversationally implicates that q to H iff:
i. S implicates that q to H.
ii. H presumes that S is observing the conversational maxims (or CooP)
when S says that p to H.
iii. If S's saying that p is S's total contribution to the conversation at
that point, then S's saying (only) that p is not consistent with the
presumption that S is observing the conversational maxims (or CooP).
iv. Only on the supposition that S thinks that q, is S's saying that p
consistent with the presumption that S is observing the conversational
maxims (or CooP).
v. S thinks (and expects H to think that S thinks) that it is within the
competence of H to work out or grasp intuitively that iv.
Since it is a necessary condition for something to be a conversational
implicature that it at least be capable of being worked out, we can
schematize H's working out in terms of this definition (Grice (1975, 70)
mentions all but (2).):
1. S said that p.
2. If S' s saying that p is his total contribution to the conversation at this
point, his saying only this is not consistent with the presumption that S
is observing the conversational maxims (or CooP).
3. There is no reason to suppose that S is not observing the conversa-
tional maxims (or CooP).
4. Only if S thinks that q, is S's saying that p consistent with the
presumption that S is observing the conversational maxims (or CooP).
5. S knows that I can figure out 4.
6. S has done nothing to block my thinking that q.
7. Therefore S intends me to think that q.
8. So S has implicated that q.
Notice that the implicature (that q) has been achieved by virtue of
special features of the context, and very much in the pattern of the
SAS.33 This sort of conversational implicature Grice calls particular-
ized conversational implicatures. As he says (1975, 56), "There is no
Issues 170
room for the idea that an implicature of this sort is normally carried by
saying that p. "
Grice coins the category of generalized conversational implicatures
for cases where saying that p would normally carry such and such an
implicature. As examples of generalized conversational implicatures,
he gives the following:
(53) a. "x is meeting a woman this evening" implicates "the woman is
not his sister, mother, wife, or close platonic friend."
b. "x went into a house yesterday and found a tortoise inside the
front door" implicates "the house was not his own."
Grice claims that sometimes, as with "I've been sitting in a car all
morning," there is no such implicature. It is an interesting question
why not, since there is an implicature in "x climbed into a car yesterday
and found a tortoise behind the seat." 34
Grice concludes (1975, 57-58) that a conversational implicature pos-
sesses certain features. (1) Since observance of the conversational
maxims is a necessary condition for calculating a conversational im-
plicature, a generalized conversational implicature can be canceled by
either explicitly or contextually opting out. (2) Since calculation of a
conversational implicature requires determination only of (i) contextual
information, (ii) background information, and (iii) what is said, but not
the manner in which it is said, then any way of saying what is said is
likely to have the same conversational implicature. Since a generalized
conversational implicature is fairly insensitive to context and back-
ground information, it should have a high degree of nondetachability.
(3) Since calculation of the conversational implicature requires prior
knowledge of what is said, the conversational implicature is not a part
of the meaning or force of what is said. (4) Since what is said may be
true and what is conversationally implicated false, the implicature is
not carried by what is said, but only by "the saying of what is said or by
'putting it that way' ." 35 And (5) in many cases the conversational im-
plicature is a disjunction. . results simply from H's concluding (or intuiting) that S's believing q is
All of Grice's examples, we might add, are clear cases of indirect
constatives.
8.4.1. Some Issues Concerning Conversational Implicature
Grice's account of the various conversational data was a significant
advance, in both rigor and insight, over previous accounts. However, it
raises a number of issues needing further work, among which are the
Philosophical and Linguistic Issues 171
following two. First, there is a problem with the mechanisms that are
supposed to generate conversational implicatures. Recall that Grice
distinguished four ways in which a maxim could be infringed: by opting
out, by violation, by clash, and by flouting. At times Grice seems to
reserve the title conversational implicature for those aroused by flout-
ing a maxim. For instance, he writes "The presence of a conversational
implicature must be capable of being worked out; for even if it can in
fact be intuitively grasped, unless the intuition is replaceable by an
argument, this implicature will not count as a conversational implica-
ture; it will be a conventional implicature" (1975, 50).
When one turns back to the working out schema, it seems that such a
working out is sufficient for flouting, since step 2 requires a maxim to be
infringed and step 5 requires H to suppose that S thinks H is aware of
this in making the inference to the implication that q. Thus, a conver-
sational implicature must at least be capable of being produced by
flouting a maxim. At other times Grice writes as if to be conversational
an implicature need depend merely on the supposition that some rele-
vant maxim (or maxims) be in effect. For instance, Grice (1975, 51)
offers the following dialogue as an example of B's implicating that
Smith has a girlfriend in New York:
(54) A: Smith doesn't seem to have a girlfriend .these days.
B. He has been paying a lot of visits to New York lately.
Part of the problem here is Grice's use of the notion of an implicature
being capable of being worked out. It seems that implicatures produced
by means of violation, clash, and flouting are all capable of being
worked out, but with flouting, the implicature is intended or intended to
be recognized as intended. 36 Since it is plausible to suppose that all
implicatures involve infringement or apparent infringement of some
maxim, we propose that the term conversational implicature be re-
served for this latter kind of case where H's inference that q is intended
to be recognized as intended. 37 When the inference to the implicatum q
required to preserve conformity to the maxims, we can call this im-
plicature simpliciter. Distinguishing these cases has the benefit that
complex inferences involving flouting a maxim need not be sought to
explain every case of implicature. (See Kempson 1975, ch. 8, for
examples, and Hamish 1976b for further discussion.)
A second problem area is generalized implicature, where "I broke a
"finger" is said to imply that the finger was the speaker's (attached).
Although Grice is inclined to think of these as conversational implica-
Issues 172

tures, it is not clear that such cases involve infringement of a maxim. Chapter Nine Indirect Acts and
But if no maxim is infringed there is nothing to work out, in which case Illocutionary
by definition they would not count as (conversational) implicatures. On Standardization
the other hand, they are not conventional implicatures either, since the
implicatum is cancelable: "I broke a finger from my finger collection."
Is there an alternative to the choice between conversational and Con-
ventional implicature that might plausibly describe these cases? We
suggest that these are simply immediate, plausible (or precedented)
inferences by H from what S has said. If this is correct then all of the
phenomena Grice has discussed under the rubric of (nonconventional)
implicature can be subsumed under some aspect of the SAS.
According to the speech act schema, all illocutionary acts but conven-
8.5. CONCLUSION
tional ones are performed with R-intentions and succeed only if those
intentions are recognized. Even literal illocutionary acts involve R-
We have contrasted what is said in the utterance of an expression with
intentions, contrary to the view that illocutionary force is part of con-
the notions of what is meant (both in the language and by the speaker),
ventional meaning. According to that view, no distinction is to be
with what is presupposed (both by the expression uttered and by the
drawn between locutionary acts and literal illocutionary acts. Although
speaker), and finally with what is implied. Our purpose was twofold: to
we allow for illocutionary force potential, as evidenced by our notions
separate (and untangle) a number of distinct concepts and to relate each
of L-compatibility and F-determinacy, we maintain that for any utter-
of them to the speech act schema.
ance to be a communicative illocutionary act, it must be issued with an
In order to specify, at line L2 of the schema, the operative meaning
R-intention. Except for conventional cases, no utterance can count as
of the expression uttered, we had to specify some type of linguistic
an illocutionary act solely in virtue of what is uttered. The speaker can
meaning (language, dialect, or idiolect). In accordance with the infer-
always fail to have an illocutionary intent or can be speaking nonliter-
ential nature of the SAS we tentatively endorsed an inferential account
ally or indirectly. Therefore, the meaning of what is uttered can at most
of meaning specification. However, an adequate conception of mean-
determine the identity of whatever literal illocutionary act is being
ing specification does not automatically yield an adequate theory of
performed, if any. It cannot determine that any such act is being per-
meaning attribution, inasmuch as it leaves open the answer to the
formed.
question of what it is for an expression e to mean .... 38 The notions of
Having recapitulated this basic feature of our position, we can pose
semantic and pragmatic presupposition have been applied in the liter-
the problem of illocutionary standardization. Certain indirect illocu-
ature to a wide variety of phenomena. However, we found that the
tionary acts do not seem to fit the SAS. These involve the use of certain
facts supporting semantic and pragmatic presupposition are better ac-
standardized sentence forms, such as the following:
counted for, respectively, by the (already needed) notions of en-
tailment and mutual contextual belief. Finally, we showed how the (1) Can you pass the salt?
information presupposed by the inferences underlying Grice's conver- (2) You might consider dropping out.
sational implicatures is already represented by aspects of the SAS. (3) I must ask you to leave.
Indeed, all of Grice's examples are cases of indirect constatives. We (4) I want to thank you for coming.
conclude from this survey that the theory presented in part I provides a
framework that will support most, if not all, of the major concepts in Although each can be used literally, (1) as a question and the others as
the study of locutionary acts and their relations to linguistic com- constatives, they are forms of sentences standardly used for acts of
munication. requesting, advising, demanding, and thanking, respectively. Because
Issues 174
their use is standardized, the hearer can determine the speaker's illo-
cutionary intent just as immediately as if a literal illocutionary act were
being performed. The process of inference spelled out in the SAS is
short-circuited: instead of having to rule out the literal intent as primary
and infer S's indirect intent, H can identify the indirect intent without
having to search for it. Standardized indirect acts are like literal acts in
that the identity of the speaker's illocutionary intent is the first candi-
date to be arrived at in the process of inference. Therein lies the prob-
lem. Is the standardization of the use of such sentence forms a matter of
meaning, in which case they must be regarded as systematically ambig-
uous? Or is standardization a matter of linguistic convention somehow
distinct from meaning? Perhaps it is neither. We find the first alterna-
tive unattractive and will argue that it multiplies meanings beyond
necessity. As for the second, we will argue that illocutionary stan-
dardization is not a case of convention, although it comes close.
9.1. THE AMBIGUITY THESIS
The ambiguity thesis is the view that sentences standardly used indi-
rectly have additional meanings. On this view the standardized use of
such sentences is not indirect after all, but literal and direct. If ambig-
uous, these sentences have two meanings which delimit two literal
uses. For example, sentence (1) could be used literally either as a
question or as a request. Since it has the surface form (interrogative) of
a question, we will designate its use as a question as its direct use and
its use as a request its indirect use. The diachronic vision (Sadock 1974,
98) behind the ambiguity thesis is that sentences like (1) were not al-
ways ambiguous and originally could be used literally and directly only
as questions. With the standardization of their indirect use as requests
they came to have a second literal meaning; in this way their secondary
literal use was preceded historically by a pattern of indirect departure
from their basic literal use.
There are two kinds of arguments for the ambiguity thesis, psycho-
logical and linguistic. The psychological argument appeals to the intro-
spective immediacy of the inference to the secondary illocutionary
intent; the linguistic argument appeals to various paraphrastic and dis-
tributional phenomena.
The primary claim underlying the psychological argument is that the
second meaning of sentences like (1) is not computed via the basic
meaning. We seem to identify the indirect intent of an utterance of
Indirect Acts 175
(1) as we would the direct intent in contexts where that is more natural.
For example, if (1) "Can you pass the salt?" is uttered at the dinner
table, it would normally be taken as a request for the salt, whereas if it
were uttered by a physical therapist to a patient recovering from polio,
it could be taken as a question. The indirect use is more common than
the direct use, but the psychological argument for the ambiguity thesis
does not depend on the indirect use being predominant. Indeed, in
some cases it is not, as in "Do you have any weapons?"
The psychological argument comports well with the introspective
consideration that the basic meaning of these standardized locutions
does not flash across one's mind in contexts where the indirect use is
the obvious one. So not only is the indirect illocutionary intent not
inferred from the direct one, the direct one does not even enter into the
determination of the indirect intent (or if it does, one is not aware of it).
Conscious disambiguation is no more necessary here than with sen-
tences like "He put his money in the bank" and "He jumped into the
water from the bank" -it is difficult (though hardly impossible) to
imagine circumstances in which either of these sentences would have to
be consciously disambiguated with respect to the word "bank."
Since sentences like (1) -( 4) do not have the superficial form of sen-
tences usually used to make requests, and so on, a proponent of the
ambiguity thesis would concede that if such a sentence did not have its
purported second meaning, its indirect intent would have to be inferred
from its direct intent. However, since the indirect use is standardized
(in virtue of an alleged second meaning), no such inference is necessary
or actually made. If the psychological argument for the ambiguity
thesis is correct, the apparent absence of inference (and of any inten-
tion that such an inference be made) can be accounted for only by
supposing that the sentences in question are ambiguous. To parry this
argument we must either demonstrate the psychological reality of the
process of inference (and of the appropriate speaker intention), even
though it is not conscious, or we must accept the claimed psychological
facts put forth in support of the argument but show that they can be
accommodated without positing second meanings. When we present
our account of illocutionary standardization in section 9.3, we will take
the latter course.
The primary claim underlying the linguistic arguments is that the
ambiguity thesis is to be incorporated in an adequate theory of a certain
range of linguistic phenomena. Sadock (1974, chs. 4-6) systematically
tries to work out such a position. He presents half a dozen arguments
Issues 176 Indirect Acts 177

against any theory that claims all cases are inferential in the way that
"My mouth is parched" or "It's really getting late" are. Some of
Sadock's arguments boil down to simplicity arguments-an inferential
theory will be more complicated than an ambiguity theory (or require
duplication of machinery) (1974, 79-82). Other arguments concern lan-
guage change (pp. 91-93) and language comparisons (pp. 93-94). The
remaining arguments (pp. 82-83, 88-91) present challenging evidence
in favor of noninferential theories.
Sadock does not want to handle all cases of illocutionary stan-
dardization in terms of force-ambiguity . But which ones? Without tests
to distinguish cases the theory would collapse into vacuity, claiming
that only those cases that can be treated as ambiguous are ambiguous.
To avoid this problem, Sadock (ch. 5) offers three sorts of test for the
existence of an idiomatic reading and thus ambiguity: cooccurrence
restrictions, paraphrasability, and transformational accessibility. Con-
sider" Spill the beans": as an idiom (a) it will not take nominal or verbal
modifiers freely ("He (*clumsily) spilled the (*green) beans"); (b) it
resists paraphrase by substitution (*"He spilled the legumes"); and (c)
it does not undergo a number of transformations-certain reductions ,
for example (*"Emest spilled the beans and Max the chickpeas").
Sadock then applies these sorts of tests to sentences like:
(b) Shouldn't you/we VP?
(c) How(s) about VP-ing?
Indirect force: (I suggest that you/we) VP.
3. Queclaratives: Does anyone VP anymore?
Indirect force: (I declare that) no one VPs anymore.
4. Pseudo-imperatives:
(a) Seek and ye shall find.
Indirect force of a generic statement: Anyone who seeks will find.
(b) Move and I'll shoot; Move or I'll shoot.
Indirect force of a threat: If you move, I'll shoot; If you don't move
I'll shoot.
(c) Eat your vegetables and I'll give you dessert.
Indirect force of an offer: If you eat vegetables, I'll give you dessert.
5. Requestions: Columbus discovered America in?
Indirect force of a disinterested request for someone to say the right
answer.
6. Tag questions:
(a) John likes beans, doesn't he?3
Force (indirect?) of a reported assumption plus a question.
(b) John likes beans, doesn't he!4
Force (indirect?) of a statement plus request for confirmation.

(5) a. Can/can't/could you VP?
b. Will/won't/would you VP?
He comes to the conclusion that "we perceive, as speakers of English
that ["Will you close the door?"]-despite its surface form-is a re-
quest on one reading-in other words, that it is actually ambiguous
between a request sense and a question sense" (1974, 108). That is to
say, these are speech act idioms. 1
Sadock then extends his theory from these whimperatives (as he calls
Since opponents of the ambiguity thesis (Searle 1975a, for instance)
have concentrated on interrogatives and declaratives used as direc-
tives, classes 1 and 2 constitute the main common data base. How good
is the case that sentences of forms such as (6) are ambiguous between a
question reading and another reading?
(6) a. Can you VP? (request for action)
b. Why don't you VP? (suggestion)
9.1.1. Wbimperatives

l
them) to various other cases of "force-ambiguity" (ch. 6). Although
some of the theoretical conclusions he draws are questionable (see
Hamish 1978), some of the data he adduces do need explaining. Sadock
deals with six kinds of cases: Can
Will SUbjUnctive} VP?
(7) .
{ negative you.
*Shall
1. Whimperatives: Can you VP? *May
Indirect force: (I request you to) VPP
Tag-imperatives (fractured whimperatives): VP, will you? The proposal is that sentences having a surface structure like (7)
Indirect force: same as for whimperatives. come via "whimperative formation" from a semantic structure like that
2. Impositives: in figure 9.1, which is also thought to underlie imperatives. The obvious
(a) Why don't you/we VP? way to argue for deriving surface structures of form (7) from such
Issues 178 Indirect Acts 179

c. *1 enjoy playing tennis, please.

d. *When will you keep score, please?5
NP On this version of the ambiguity theory, the verb denoting the action
requested will be in VP 2 of S2 in figure 9.1, and so is "formally" locata-
ble. Thus, the data (8) and (8') provide some support for deriving sen-
v NP tences of form (7) from structures like that in figure 9.1.
Second, only the tagged ("fractured") forms of (7) are grammatical
with "please" either preverbally or postsententially:

(7')
(Please) VP,
can
will I SUbjUnctiVe}
.
{ negatIve you, ( I)
pease

I REQUEST YOU
Figure 9.1 Semantic structure underlying (7)
semantic structures is to find evidence that these sentences (in their
indirect use) behave as imperatives. Another strategy is to find evi-
dence that these sentences (or transforms thereof) do not (in their
indirect use) behave as interrogatives used to question. Sadock offers
three main sorts of evidence in favor of the ambiguity thesis.
First, there are considerations about the distribution of words like
"please" and "kindly." In particular, "please" seems to occur pre-
verbally and without a pause just when the verb denotes a requested
action (pp. 104, 124).
(8) a. Please close the door!
b. Passengers will please not lean out the windows!
c. I (*please) enjoy (*please) playing tennis.
d. When will you (*please) keep score?
Whimperatives of the form (7) conform to the generalization in (9):
(9) WilVwon't/would } I VP?
Can/can't/could you pease . tive. Thus, fractured whimperatives do not behave like interrogatives
Furthermore, if we posit an optional rule that moves "please" from
preverbal to postsentential position, we can account for the following
analogs of (8a) -(8d):
(8') a. Close the door, please!
b. Passengers will not lean out the windows, please!
1
*shall
*may
Arule relating (7) to (7') would account for these facts, but notice that if
(7') is a request, then unless the transformation changes meaning, (7)
must have a request reading as well. What evidence is there that sen-
tences of form (7') are requests and not questions? They do not behave
like questions in at least four respects. Consider the following (from
Sadock 1974, 112):
(10) a. When will you wash the car, or don't you know?
b. *Wash the car, will you, or don't you know?
(11) a. When will you wash the car, by any chance?
b. *Wash the car, by any chance, will you?
(12) a. Tell me, when will you wash the car?
b. *Tell me, wash the car, will you?
(13) a. When will you wash the car, and when will you do the dishes?
b. *When will you wash the car, and I'll do the dishes.
c. Wash the car, will you, and I'll do the dishes.
In cases (10) -(12) some type of expression that cooccurs with clear
cases of interrogatives used to question does not cooccur with frac-
tured whimperatives. In case (13) only a question can be conjoined to a
question, but a nonquestion can be conjoined to a fractured whimpera-
used to question; since they clearly can be used to request (action), if
sentences of form (7') come from (7) and transformations do not change
meaning, then (10) -(13) constitute evidence for deriving sentences of
form (7) from structures like that in figure 9.1.
Although these data support the ambiguity thesis, there are prob-
 Indirect Acts 181
Issues 180

lems. For instance, one can conjoin a nonquestion to a question if the

nonquestion relates to the question in some way:
NP
(13 ') b . When will you was h the car, an d { be careful how you answer .
I want to know quickly.
Just how this restriction might be stated formally is another problem for v
this proposal.
Third, certain expressions cooccur with (or in) clear cases of impera-
tives but do not cooccur with (or in) interrogatives (pp. 104-105):
(14) a. Wash the car, someone!
b. *When will you wash the car, someone?6
c. Will you wash the car, someone?
(15) a. Buzz off!
I SUGGEST YOU
b. *When will you buzz off?7
c. Will you buzz off! Figure 9.2 Semantic structure underlying (17)
(16) a. Since I haven't finished these exams, (please) start dinner!
h. *Since I haven't finished these exams, (please) when will you (17) a. Why don't you VP?
start dinner? b. Shouldn't y()U VP?
c. Since I haven't finished these exams, will you start dinner?
Furthermore, if we supposed that surface structures like (18)9 can be
In each case, the vocative "someone," the idiomatic imperative, and derived from (17) by a version of fracturing-
the reason adverbial whimperatives are behaving like imperatives and
not like questions. But can the ambiguity theorist really explain all of (18) a. VP, why don't you?
these (putative) facts? For instance, the connection between the reason b. VP, shouldn't you?
adverbial and the underlying REQUEST is fairly indirect; it provides a
-and that forms like (19) can be derived from (17a) by you+tense
reason for requesting only because we take into account the act being
deletion-
requested as well as general background information to the effect that it
is hard to grade exams and cook dinner simultaneously: (19) Why not VP?
(16') c. ?Since I'm chewing gum, will you start dinner? - we could account for the similarity in meaning and force among
these sentences as well as for a variety of syntactic facts, such as that
Sentences like (8)-(16) provide some, though not overwhelming,
sentences like (19) take only second-person reflexives, as in (20):
evidence for the ambiguity thesis as regards whimperatives. At least

~~~;::~es! ?
these sentences constitute interesting data that should be explained, or
explained away.
(20) Why not wash *himself
9.1.2. Impositives8

Sadock (p. 118) proposes that sentences having surface structures as in

1 *myself
*themselves
(It seems to us that "Why not wash ourselves?" is acceptable too.) Is
(17) come, on one reading, from semantic structures like that in figure
there any evidence that these impositives are (semantically) sugges-
9.2.
Issues 182 Indirect Acts 183

tions (pp. 119-120) and neither questions (pp. 114-115) nor whimpera- (25) a. Why don't we go to the movies?
tives (p. 116)? Evidence that they are suggestions is very weak and that b. Won't you/*we go to the movies?
they are neither questions nor whimperatives is only a bit stronger.
Impositives are suggestions. First, it is claimed that with suggestions Moreover, these impositives cannot be fractured:
it is the case both that one can assent with "OK" (" Sure," "Alright," (26) *Go to the movies, why don't we?
"You bet," and so on)10 and that one must refuse with "No" ("Nope,"
"Nab") plus a reasonfor refusing if one is to be accommodating (raise Second, impositives resist preverbal "please":
no conversational animosities): (27) ?Why don't you please go to the movies?
(21) A: Let's go to the movies tonight! Finally, impositives may require slightly different negative responses
B: OK. than whimperatives (the data are tricky):
B': No, I have to study.
B": *No. (28) a. A: Would (could, etc.) you move over a little?
B: I can't, ...
The same is true of impositives: B: No, I can't, ...
(22) A: Why don't you go to the movies? (Shouldn't you go to the b. A: Why don't you move over a little?
movies?) B: I can't, ...
B: OK. (But *Yes.) B: ??No, I can't, ...
B': No, I have to study. Impositives are not questions. First, questions take questioning of
B": *No. their epistemic felicity conditions, but impositives do not:
Unfortunately, the line between these cases and clearly imperatival (29) When (at what time) are you going to the movies, or don't you
cases is very delicate: know?
(23) A: Go to a movie tonight! (30) *Why don't you (please) go to the movies, or don't you know?
B: OK. Second, impositives can, and questions cannot, take a nonquestion
B': No, I have to study. conjunct unrelated to the first question:
B": ?No.
(31) Why don't you go to the movies, and I will too.
Second, Sadock (1974, 120) notes that suggestions can take the post- (32) *When (at what time) are you going to the movies, and I will too.
sentential tag "OK?" as do whimpositives (but not "shouldn't" im-
positives) : Thus there does exist some distributional evidence for treating
whimperatives and impositives as having an underlying structure dis-
(24) a. Let's go to the movies tonight, OK? tinct in illocutionary type from their surface structure. Even if the case
b. Why don't we go to the movies tonight, OK? for the ambiguity thesis is not particularly strong, at least it represents
c. *Shouldn't we go to the movies tonight, OK? an effort to explain these data. An adequate alternative must explain
What Sadock does not note is that this seems to be true of imperatives them better.
as well:
9.2. THE CONVENTIONALITY THESIS
(24) d. Go to the movies, OK?
Impositives are not whimperatives (nor imperatives). First, imposi- Although the ambiguity thesis does account for the indirect uses of
tives do and whimperatives do not take an inclusive first-person plural sentences like (1a)-(1d), a viable alternative to it would have the virtue
subject: 11 of not multiplying meanings beyond necessity. Of course, if the psy-
Issues 184 Indirect Acts 185

chological and the linguistic arguments for it were sound, the thesis tion like interrogatives (in their direct use). (See Green 1975,109-114,
would not be guilty of multiplying meanings beyond necessity - the 137-138; Sadock 1974, 113.)
arguments would necessitate the multiplication. 12 These arguments are In the first place, conversationally felicitous responses to whimpera-
not conclusive, but we should not reject the ambiguity thesis unless We tives pattern like responses to interrogatives, not like responses to
can produce a viable alternative. One possibility is that the indirect Use imperatives, in that whimperatives require a verbal response whether
relies on illocutionary conventions. We will call this the conventionality or not they are complied with, imperatives only if they are not:
thesis.
(33) A: Do you have the ability to pass the salt?
Being able to produce and to perceive standardized illocutions seems A': Can you pass the salt?
to be part of a speaker's linguistic knowledge. Even if it is not knowl-
B: *salt-passing
edge of the meaning of sentences of certain forms, it is knowledge
(34) A: Pass the salt!
about their use. On the conventionality thesis this is knowledge that the B: (optional: Sure) salt-passing
utterance of a sentence of a certain form literally used to perform one
sort of illocutionary act counts as the performance of some other sort of Furthermore, "Yes" without a title or vocative ("Yes, sir," "Yes,
illocutionary act.13 On this view a speaker's knowledge ofillocutionary ma'am") is a bit odd as a response to an imperative, though "OK,"
conventions is not part of his linguistic competence proper, and need "Sure," and so on, are all right.
not be captured in a descriptively adequate grammar, although one
(35) A: Pass the salt!
could, as a matter of terminological preference, include such knowl-
B: ?Yes.
edge as part of linguistic competence. Whether or not one does, on the
conventionality thesis the existence of illocutionary conventions ac- Second, Sadock (p. 113) has noted that directives can be reported
counts for the fact that standardized illocutionary acts can be per- using "ask" and "tell," but questions prefer "ask" and imperatives
formed in the indirect but immediately identifiable way that they are. prefer "tell":
Since these conventions involve mutual belief that certain sorts of ut-
(36) A: When are you going to VP?
terances count as certain sorts of illocutionary acts, speakers and hear-
B: S asked H when ...
ers can omit the intermediate steps in the SAS for indirect acts. These
B': *S told H when ...
conventions serve to bypass the inference, and its conclusion is
(37) A: Take out the garbage!
reached without further ado. As Searle (l975a, 73) says, the hearer
B: S told H to take out the garbage.
"simply hears it as a request." It is clear, then, how the conventionality
B': *S asked H to take out the garbage.
thesis could use introspective evidence for its position and so circum-
vent the psychological argument for the ambiguity thesis. Whimperatives appear to pattern with questions:
Linguistic considerations provide evidence against the ambiguity
thesis as well as positive evidence for the conventionality thesis. (38) A: Can you VP?
B: S asked H to VP.
9.2.1. Against the Ambiguity Thesis B': *S told H to VP.
One complication is that "ask" and "tell" appear to be sensitive not
Evidence that whimperatives and impositives function as questions in just to the form of the directive, but to its politeness:
addition to functioning as requests or suggestions is evidence against
ambiguity,14 as is evidence that their indirect force is not idiomatic. (39) A: Please VP.
B: S asked H to VP.
Whimperatives (40) A: Shut up, will you? Will you shut up!
There is some evidence that whimperatives (in their indirect use) func- B: S told H to shut up.
I
Issues 186 Indirect Acts 187

If this is right, it suggests that one should explain (36) and (37) not (44) a. Can you please VP?
directly in terms of the form of A, but in terms of the politeness of the b. *Are you able to please VP?
form. Imperatives may be a less polite form, ceteris paribus, than inter- (45) a. Will you please VP?
rogatives and this may be what is reflected by the distribution of" ask" b. *Are you going to please VP?
and "tell." Since politeness is such a tricky notion, we do not put much
It is just this that suggests that special illocutionary conventions might
weight on this observation.
be operative. We will return to this point after discussing impositives.
In sum, the data suggest that whimperatives pattern in part like
questions, in requiring verbal responses, and in part like imperatives, in
requiring action for compliance. The conventionality thesis, which has Impositives
There is similar evidence that impositives also function (in part) as
them being both, is better able to accommodate such facts than the
questions and not as direct suggestions. Impositives (like whimpera-
ambiguity thesis is.
tives) allow for a literal and direct question response:
There is evidence that whimperatives do not function as idioms
(Searle, 1975a, 68-69). First, they can be answered like questions: (46) a. A: Why don't you VP?
B: Because I am tired.
(41) A: Could you be quiet?
b. A: I suggest you VP.
B: I could, but I won't.
B: *Because I am tired.
Such dialogues seem "smart-alecky" (Green 1975); this is true of
Again, these are smart-alecky, and the reason is the same as before-
whimperatives in a way that it is not true of answers to (mere) ques-
only part of the speaker's communicative intent has been cooperated
tions. But that does not tell against whimperatives functioning at least
as questions. In fact, there is a specific reason why such responses are with.
smart-alecky: when it is obvious to S that H realizes that S in ques-
9.2.2. For the Conventionality Thesis
tioning is also requesting action, H's responding appropriately to just
the direct illocutionary intent is seen as linguistically defensible ("You
Do proponents of the conventionality thesis have data to support their
asked me if I could") but socially uncooperative. Following the letter
claim as well as explanations for the evidence supporting the ambiguity
but not the spirit of an utterance is just the sort of thing we deem
thesis? If we tum to Searle (1975a)-the most sophisticated articula-
smart-alecky. Rather than detracting from the conventionalist's case,
tion of this position so far-we find the theory wanting. Searle notes
Green's observation enhances it.
just two of the various types of syntactic data that support the ambigu-
Another consideration is that word-for-word translation often pro-
ity thesis. First, he observes thaf"please" can occur postsententially
duces expressions with the same indirect speech act potential (see
with certain nonimperative surface forms such as
Searle 1975a, 68):
(47) a. I want you to stop making that noise, please. (p. 68)
(42) a. Could you give me the salt?
b. Could (can) you lend me a dollar, please? (pp. 68, 75)
b. Pourriez-vous me donner Ie sel?
but not with other forms:
This is not the case with most idioms:
(48) a. *Do you desire to do A, please? (p. 75)
(43) a. How are you? *Comment etes-vous?
b. *Are you able to do A, please? (p. 75)
b. ~a va? *It goes?
Second, Searle notes that "Why not ... ?" impositives with a sec-
Sadock (p. 90) observes that, within English, whimperatives resist
ond-person direct object require a reflexive ("Why not wash (*you)
paraphrase, preserving indirect speech act potential:
yourself?") just like imperatives ("Wash (*you) yourself!") and un-
Issues 188
like other interrogatives ("Why won't he wash (*yourself) you?") and
declaratives ("He will wash (*yourself) you"). How does Searle ac-
count for these facts? He offers no account of the reflexive-impositive
data (p. 78).
As for the "please" data, Searle rightly says that "please" "ex-
plicitly and literally marks the primary illocutionary point of the utter-
ance as directive" (p. 68). But what is it to mark the point as directive?
It is not clear whether "please" is itself a directive force-indicating
device or whether it merely cooccurs with something that is. In either
case, why are (48a, b) so bad? Searle's answer (p. 76) is in two stages.
He claims first:
Certain forms will tend to become conventionally established as the
standard idiomatic forms for indirect speech acts. While keeping their
literal meanings they will acquire conventional uses . . . there can be
conventions of usage that are not meaning conventions.
Second, he claims that conversations are governed by the maxim:
Speak idiomatically unless there is some special reason not to.
How do these observations add up to an explanation of the "please"
data? In Searle's words,
In order to be a plausible candidate at all for use as an indirect speech
act, a sentence has to be idiomatic. But within the class of idiomatic
sentences, some forms tend to become entrenched as conventional
devices for indirect speech acts. In the case of directives, in which
politeness is the chief motivation for the indirect forms, certain forms
are conventionally used as polite requests. (p. 77)
It is not clear to us how these observations explain the data. Part of
the problem is the obscurity of the notions of speaking idiomatically
(which is not to amount to speaking with idioms), and conventions of
use (which are not conventions of meaning). We suspect that with
regard to the first point Searle probably means the maxim to read
"speak colloquially" rather than "speak idiomatically." Surely no
maxim of conversation dictates that one should say (49a) instead of
(49b):
(49) a. Joan's old man kicked the bucket.
b. Joan's husband died.
However, "speak colloquially" is almost as empty as "speak idiomati-
cally" is obscure (if it has nothing to do with using idioms). To speak
Indirect Acts 189
colloquially is just to speak as people speak, rather than as they write.
Such a maxim can have little utility in explaining the "please" data.
As for the second point, it is not clear what Searle takes conventions
of use to be, as opposed to conventions of meaning. If by conventions
he means something like customs, the explanation of the data would
presumably reduce to this:
1. It is customary to request by using certain declarative and interroga-
tive forms (if not using an imperative).
2. Conversations are governed by the maxim "Speak colloquially":
speak as people normally speak unless there are reasons to the con-
trary.
3. Postsentential and preverbal "please" indicates that the speaker is
intending to request.
4. It is not customary to request using sentences of form (48) (without
"please"), but it is customary to request with (47).
5. So (48a,b) are odd, but (47a,b) are not.
By this account the oddity of (48) reduces to a statistical fact about the
frequency of requests, but surely the oddity does not consist in this.
Nor is it clear exactly how it follows that (48a, b) are odd. To get an
explanation out of 1-4 we need an account of what it is customary to
request with. Given such an account, how would we explain these
judgments of oddity by fluent speakers? What is the connection be-
tween a statistical fact about usage and these psychological states of
speakers? There is no stated connection and so no explanation. Even if
Searle could work up an explanation of these judgments, he could still
not account for the following distributional facts concerning "please":
(50) a. How old (*please) are you, (please)?
b. Are you (*please) able to do A, (please)?
c. Can you (please) do A, (please)?
Since taking conventions to be customs does not seem to explain the
facts, let us try construing conventions as we did in section 6.1. Are
there conventions of use in this interpretation? They would take the
following form, if we let T range over sentence forms (such as "Can
you ... ," "Will you ... "):
Illocutionary Convention (IC): There is an (illocutionary) convention in
group G for F-ing in uttering (a sentence ofform) T (in context C) if and
only if:
Issues 190 Indirect Acts 191

i. It is MB-ed in G that whenever a member of Gutters Tin C, he is the way Searle sets up the problem of indirect speech acts: "How is it
F-ing, and possible for the speaker to say one thing and mean that but also to mean
ii. Uttering T in C counts as F-ing only because it is MB-ed in G to something else . . . how is it possible for the hearer to understand the
count as such. indirect speech act when the sentence he hears and understands means , ,I

something else?" (1975a, 60) On the one hand, if in performing indirect

An example of one such sentence type is (1), repeated here:
illocutionary acts the speaker means (and intends it to be recognized
(1) Can you pass the salt? that he means) to be speaking directly as well, then the hearer cannot
understand the speaker and just hear it as a request. If, on the other
To utter (1) is to ask whether the hearer is able to pass the salt. Suppose
hand, there were conventions of use that would allow the speaker to
the situation is one in which to utter (1) with merely its literally deter-
perform the indirect act conventionally and so allow it to be heard as a
mined force (to ask the hearer whether he can pass the salt) violates the
request only, then it does not conform to putative facts 7 and 8, nor to
presumption of relevance (RE) and thereby the communicative pre-
the statement of the problem in the first place. It would no longer be an
sumption (CP). Then if there is a convention to the effect that to utter a
indirect act, by Searle's own definition.
sentence of the form "Can you ... ?" counts as a request, the speaker
One way out of this dilemma would be to make a sharper division
might expect to achieve his illocutionary intent. 15 But are there such
between those cases governed by some additional conventions of use
illocutionary conventions? We think not.
(where such conventions need to be explicated in some as yet unknown
Utterances of the locutions in question can count as the performance
way) and those cases that are not. Although Searle distinguishes be-
of the specified type of illocutionary act without being mutually be-
tween the forms (51a) and (SIb) with respect to conventions of use, he
lieved to count as such. In accordance with the SAS, the hearer can
does not distinguish between (SIb) and (SIc) in this regard:
identify the speaker's illocutionary intent without recognizing any con-
vention or participating in the mutual belief falling under it. Relying on (51) a. Is it the case that you presently are capable of VP-ing?
the communicative presumption, he can make the inference in the b. Can you VP?
usual manner. Thus, satisfaction of clause (ii) of the definition of illocu- c. Are you able to VP?
tionary conventions (IC) is not necessary for the performance of the
If one supposes that conventions of use govern (SIb) only, then one
illocutionary act of F-ing in C when one utters a sentence like (1).
could say that in such a case one hears it as a request in virtue of these
Finally, there seems to be a prima facie conflict between the psy-
conventions. However, one would still have to deny putative facts 7
chological argument and the linguistic argument for the conventionality
and 8 for these cases, and these conventions of use would have to be
thesis. Recall that psychologically the hearer "just hears it as a re-
stated in such a way as to allow for the explanation of the relevant
quest" - he understands just one act to have been performed. Yet
linguistic data. Clearly the work has just begun for the conventionality
Searle also wants to subscribe to the "putative facts" he numbers as 7
thesis.
and 8 (1975a, 69-70):
[Fact 7] In cases where these sentences [like "Can you pass the salt?"] 9.2.3. Ambiguity versus Conventionality
are uttered as requests, they still have their literal meaning and are
uttered with and as having that literal meaning. The main strength of the ambiguity thesis is that it has outlined how
[Fact 8] It is a consequence of fact 7 that when one of these sentences is certain syntactic relationships might be formalized. However, it has
uttered with the primary illocutionary point of a directive, the literal not yet produced an account of most of the syntactic data adduced in its
illocutionary act is also performed. own support- no actual derivations using antecedently motivated
rules are given. Of course, anyone familiar with how a transformational
These facts suggest that two acts are being performed. The convention- grammar accounts for various linguistic facts might be tempted to sup-
ality thesis cannot have it both ways. The problem is exacerbated by pose that explaining these facts will just be more of the same. But that
Issues 192
Indirect Acts 193

is a mistake for two reasons. First, rules that seem plausible in isolation
In rejecting the conventionality thesis, we saw that clause (ii) of the
can generate serious problems when one attempts to make them
definition of illocutionary conventions (IC) is unnecessary for the ef-
work together. 16 Second, much of the data involve sentence-final co-
fectiveness of a standardized illocution. The reason is that the illocu-
occurrence restrictions of force (having little to do with traditional
tion would be effective even if not standardized, since the hearer could,
transformations) which are susceptible to alternative explanations in
if necessary, infer the speaker's illocutionary intent from what is ut-
terms of speech acts. After all, there is no reason to suppose every
tered and the relevant mutual contextual beliefs, as with any indirect
oddity induced by putting words together must be explained as a
illocutionary act. Indeed, it is assumed on all si~es that these stan-
grammatical oddity. The main weakness of the ambiguity thesis (in its
dardized illocutions become standardized through use over time. Only
present form) is its failure to show how, in general, indirect acts are
by accumulating precedent for indirect use do such sentences come to
related to direct acts. This would involve, in part, a theory of speech
be standardized, and their being standardized consists in whatever it
acts rich enough to support generalizations concerning alternative
takes for the SAS to be short-circuited. What this amounts to is the
ways of performing illocutionary acts, where the expressions used do
satisfaction of clause (i) of IC. Putting these observations together, we
not (or need not) bear any derivational relation to each other.
offer the following preliminary definition of illocutionary standardiza-
The main str~\1gth of the conventionality thesis is that it is a part of an
tion: 17
interesting theory of speech acts, a theory that contains generalizations
relating indirect to direct acts. Its main weakness is the implausibility of Illocutionary Standardization (1): Tis standardly used to Fin G if and 'I
illocutionary conventions and of how they could account for the lin- only if:
guistic facts. A wide variety of relations between sentences and illocu- i. It is MB-ed in G that whenever a member of Gutters T in contexts in
tionary acts can constitute their indirect illocutionary act potential. which it would violate the CP to utter T with (merely) its literally
Clearly some are figured 'out conversationally, at least the first time determined force (F*-ing), his illocutionary intent is to F, and
they are encountered. As remarked in section 4.3, these seem to fall ii. The mutual belief in (i) is (nonvacuously) true.
neatly under the SAS. Other cases seem to be moderately plausible
Without the stipulation in (ii) that the mutual belief in (i) be true, it
cases of speech act idioms: Searle (1975a) offers "How about ... ?"
would not be the case that T is in fact standardly used to F, even if it is
used to suggest or recommend. In this and similar cases the force is
mutually believed to be so used. Of course, it is highly unlikely that
learned in the way that other idiomatic meanings are learned, as a unit.
such a mutual belief could arise without being true. The "in contexts"
In between these two extremes lie the problematic cases like "Can you
stipulation indicates that the situation is just the sort in which the
pass the salt?" which have dominated the literature. We think these are
utterance of T, if T were not standardized, would be readily identifiable
cases of standardized force, and we tum now to our account of them.
as having the illocutionary intent of F-ing. Thus, if sentence (1) "Can
you pass the salt?" were not standardized as a request, then in a con-
9.3. THE STANDARDIZATION THESIS
text where the hearer could not reasonably attribute to the speaker the
illocutionary intent of asking him if he has the ability to pass the salt he
Although it is a mistake to think of illocutionary standardization as a
might be able to infer (following the SAS) that the speaker is requesting
matter of convention, the conventionality thesis is on the right track.
him to pass the salt. The point of the definition is to capture what
Without multiplying meanings, it meets the psychological argument for
short-circuits the SAS: the utterance and the context activate the
the ambiguity thesis by providing a picture of how the SAS might be
mutual belief that the speaker and the hearer share, so that the speaker
short-circuited: the mutual beliefs that constitute illocutionary con-
can reasonably intend, and the hearer can recognize him as intending,
ventions enable the hearer to infer the speaker's indirect intent im-
to be F-ing. The intermediate steps of the SAS are thereby skipped.
mediately, without going through the usual working-out process.
It might be objected that our formulation of the nature of illocution-
Fortunately, it is possible to formulate a concept of illocutionary stan-
ary standardization is psychologically implausible-that people do not
dardization that does not require illocutionary conventions, even
in fact have the mutual belief mentioned in our definition. Indeed, they
though it does involve mutual beliefs that short-circuit the SAS.
do not individually believe, much less mutually believe, that whenever
Issues 194
certain sentences are uttered in contexts where their having (merely)
their literally determined force would violate the CP, they are used
with some other illocutionary intent. The objection, then, is that such a
belief would require that people have certain theoretical concepts that
they do not seem to have, such as the concept of literally determined
force and the concept of the communicative presumption.
We can concede the point of the psychological objection and refor-
mulate the definition of illocutionary standardization in a psychologi-
cally more realistic way:
Illocutionary Standardization (2): Tis standardly used to F in G if and
only if:
i. Whenever a member of Gutters T in contexts in which it would
violate the CP to utter T with (merely) its literally determined force, it
is MB-ed between Sand H that S's illocutionary intent in uttering T is
to F, and
ii. Whenever a member of Gutters T in such contexts, his illocutionary
intent is to F.
This formulation places the objectionable material outside the scope of
the mutual belief. Instead, the relevant context is specified and the
mutual belief that obtains therein is a contextual mutual belief between
Sand H rather than a standing mutual belief in G. Unfortunately, it is
not clear how this MB in (i) comes into existence in the context, nor
what its status is. Also (ii) seems too strong. We could opt for a standing
mutual belief in (i), namely:
i. It is MB-ed in G that whenever a member of Gutters T, his illocu-
tionary intent is to F.
But this mutual belief would be patently false. However, considering
that the sentences with illocutionary standardization are not ordinarily
used with their literally determined force, perhaps there is a weaker
mutual belief, replacing "whenever" by "generally," which would be
true:
i. It is MB-ed in G that generally when a member of Gutters T, his
illocutionary intent is to F.
What we need to do now is to formulate the clause that contextually
activates this standing MB. Accordingly, we offer the following as our
definition of illocutionary standardization:
Indirect Acts 195
Illocutionary Standardization (IS): Tis standardly used to F in G if and
only if:
i. It is MB-ed in G that generally when a member of Gutters T, his
illocutionary intent is to F, and
ii. Generally when a member of Gutters T in a context in which it
would violate the CP to utter T with (merely) its literally determined
force, his illocutionary intent is to F.18
This definition has the virtue of realistically ascribing a true mutual
belief to people, while at the same time specifying the conditions in
which the sentences in question have their standardized indirect use. It
is clear that if members of G have merely the mutual belief specified in
clause (i), this mutual belief will be activated whenever T is uttered,
thereby short-circuiting the SAS. The inference using what is mutually
believed is generally made in the appropriate contexts, since T is gen-
erally uttered only in such contexts. Therefore, S can reasonably in-
tendH to make, andH can reasonably make, a correct inference to S's
illocutionary intent.
9.4. INFELICITY AND STANDARDIZED INDIRECTION
A consideration seeming to favor the ambiguity thesis is that most
standardized illocutions appear not to be used with their literally de-
termined force at all. 19 In using "Can you pass the salt?" as a request,
one does not seem to be asking the hearer if he is able to pass the salt.
Presumably it is mutually believed that the hearer has this ability;
therefore the speaker has no reason to inquire about the matter. In-
deed, in contexts where the hearer's ability is genuinely at issue, a
literal question about that ability is usually not taken or intended as an
indirect request. But where the hearer's ability is not at issue, it would
seem that in saying "Can you pass the salt?" the speaker is not inquir-
ing about the hearer's ability at all, and therefore that he is not indi-
rectly requesting the hearer to pass the salt, contrary to what our
account of standardization requires.
However, an illocutionary act can be performed, and an attitude can
be expressed, despite obvious insincerity on the part of the speaker. 20
In particular, one can ask a question even if it is mutually believed that
one does not want to know the answer and does not intend the hearer to
provide one: S can still express this desire and this intention and
thereby ask H a question. We are suggesting, then, that an utterance of
 Issues 196
Indirect Acts 197

"Can you pass the salt?" as a request is literally an act of asking a

can be (R-intended to be) taken indirectly. Obvious pointlessness is
question, albeit with obvious insincerity. Indeed, S relies on H's rec-
another.
ognition of his (S's) obvious insincerity to convey his indirect request
What is true of "can" applies also to "could," as in the forms "Could
since otherwise, under the circumstances, there would be no identifi~
you A?" and "You could A." Although "could" in these contexts is
able reason for his utterance.
taken by some (Searle 1975a; Green 1972) as the subjunctive form of
It is often remarked that the point of requesting indirectly is polite-
"can" (in other contexts it is the past tense of "can"), we see no reason
ness. Given the social supposition that it is impolite to impose on peo-
to dispute Webster's construal of it as a polite form of the present tense
ple and in particular to tell them to do things, to request indirectly is to
of" can." After all, if" could" is really a sUbjunctive in these sentence
"a~k without asking," or to ask without explicitly asking-that is, by
forms, what are the suppressed conditions? Besides, as Searle and
domg somethin~ else and letting the request be implied. One is asking,
Green both note, supplying a condition detracts from the indirect re-
of course, but m such a way that one gives the appearance, albeit
questive force: "Could you be a little more quiet if I asked you to?" To
transparent, of not asking. Instead, one gives the appearance of pos-
support our interpretation, we note that the simple present "could" can
ing a question. The reason that a so-called whimperative works as a
be used to make a literal statement, without there being an expressed or
request is that the obviously insincere question it involves makes
suppressed condition: "You could solve that problem easily."
refer~~ce to t~e desired action by H, or at least to some obvious pre-
As Searle points out, forms like "Are you able to A?" and "You are
condltlon for It, thereby enabling H to identify the attitudes whose
able to A" are not standardly used as indirect requests, and he suggests
expression makes the utterance a request. In this way whimperatives
the reason is that they are not as idiomatic as the other forms. Since we
bring up the issue of H' s action without explicitly telling H to perform
have rejected Searle's maxim of idiomaticity, we offer a different ex-
it.
planation. Comparing "Are you able to pass the salt?" with "Can you
Let us consider the illocutionary intents-the expressed attitudes-
:1
pass the salt?" it seems that the use of "able" somehow focuses the
of various standardized forms, keeping in mind that one means for
hearer's attention on the question (or statement) of his ability, rather
performing an indirect illocutionary act is to perform a direct one with
than on the action itself.
obvious insincerity, so that the hearer is induced to seek some other
expressed attitude that could be genuinely attributed to the speaker.
9.4.2. Will, Would
9.4.1. Can, Could
An utterance of a sentence of the form "Will you A?" is a literal
question, but under the circumstances in which such a question is used
Utterances of sentences of the form "Can you A?" are literal questions
as an indirect request, were no request being made, the answer to this
about the hearer's ability to A and indirect requests that he A.21 Such
question would be negative. However, if H were to take the utterance
an utterance is obviously insincere as a question if S obviously knows
merely as a literal question and respond "No, what made you think I
the answer (and is not, for instance, testing H) or if it is obvious that S
was going to A?" clearly he would have failed to see the point of the
is not interested in the answer per se.
utterance as an indirect request. This point is (R-intended) to be in-
Utterances of sentences of the form "You canA-" can also be used as
ferred from the fact that it was obvious that H was not about to A.
requests for H to A. Such an utterance is literally a statement, but
Similarly, the negative form "Won't you A?" is used as an indirect
where it is obvious that Sand H know that H can A or where it is
request when it is expected that H won't A and recognized that for
obvious that Sand H are not interested in whether H can A H can
informational purposes the question needn't be asked.
infer that the utterance is intended as a request. 22 There is 'nothing
As with "could" we do not construe "would" in the contexts "Would
insincere about S's statement, but in the circumstances it is obvious
youA?" and "Wouldn't you A?" as the subjunctive form of "will." On
that there is not sufficient reason for S to be making merely the state-
that interpretation there is again the problem of finding the suppressed
ment. Obvious insincerity is only one way in which a literal utterance
condition and the fact that if it is supplied ("Would you leave if I asked
 Issues 198
you to?"), the utterance doesn't have the force of a request. We suggest
that "would" in the contexts here is a simple present tense verb of
willingness,just as it is in "I wouldn't do that myself." Accordingly, to
ask about H's willingness to A is, when the answer is obviously nega-
tive, a literal question intended as an indirect request.
9.4.3. Must, Ought, Should
When His A-ing, questions of the form, "Must/ought/should you A?"
are standardly used as requests for H not to (continue to) A. The
answer to these literal questions is obviously negative, so if there is to
be a reasonably attributable intent to such an utterance, it must be to
request H not to A any further. The same point applies, mutatis mu-
tandis, to negative literal questions of the form "Mustn't/oughtn't/
shouldn't you A?"23
Statements of the form "You must/ought/should A" are literal state-
ments indicating that there is a strong reason for H to A. In contexts
where it is obvious that this reason is S's desire, it may be inferred that
S is requesting H to A. Statements of S's desire ("I want you to A,"
"I'd like you to A") function as indirect requests by actually specifying
one such reason.
1'1'
I,
9.5. STANDARDIZATION AND PARAGRAMMATIC FACTS
In the standardization thesis, the requestive use of certain interrogative
forms short-circuits the SAS, the hearer identifying the speaker's re-
questive illocutionary intent without having to identify the literal intent
of questioning. He does this by relying on the precedent for the form's
being used requestively. This precedent and the hearer's unthinking
reliance on it depend on the form's being specifiable in some syntacti-
cally determinate way, and it must be such that the connection between
the literal force/content and the indirect force/content is simple and
systematic. Then both the existence and the identity of the indirect
illocutionary intent can be immediately inferred by the hearer.
The standardization thesis does justice to the psychological facts
without assigning additional meanings to the forms in question. How-
ever, there are also linguistic facts (or seeming facts) to account for,
and Sadock (1974) relies on such facts in developing his version of the
ambiguity thesis. One such fact is the preverbal occurrence of "please"
in interrogative forms like (52) and (53) when used as requests.
Indirect Acts 199
(52) Can you please pass the salt?
(53) Will you please pass the salt?
Without "please" these sentences can be used as questions; with
"please" they can be used only as requests. To account for this fact, as
well as to account for the requestive use of such sentences (with or
without "please"), Sadock supposes that they have an underlying im-
peratival form and that they are only interrogative on the surface. On
this supposition it is easy to see why the "please" can occur pre-
verbally, just as in imperative sentences.
The acceptability of "please" in these contexts presents a problem
for the standardization thesis. If (52) and (53) without "please" are
unambiguously interrogatives, how can "please" acceptably occur in
them? We suggest that sentences like (52) and (53) in which "please"
occurs, are not fully grammatical, although their recurrent use makes
them seem perfectly acceptable;24 they are not grammatical interroga-
tives and cannot be used as literal questions. Although "please" can be
paraphrased as "if you please," "ifit pleases you," "be so good/kind as
to," or in some such way-any of these paraphrases seems plausible to
us-it can occur only in sentences used to make requests (or com-
mands) not in sentences used to ask questions or to make statements.
For example, it seems that (54) and (55) are clearly not grammatical. 25
(54) *Why do you please pass the salt?
(55) *You never please pass the salt.
So it appears that (52) and (53) are grammatical only if used as requests,
but this suggests that without "please" they have two literal readings,
one interrogative and one imperative. To avoid this consequence while
allowing that "please" can be used, in the relevant sense, only to make
requests (or commands), we propose that (52) and (53) are not gram-
matical sentences and should not be generated by a grammar of
English. Instead, they are examples of the phenomenon of syntactic
liberty (see section 10.4), ungrammatical but usable sentences that are
perfectly acceptable to fluent speakers.
Of course, it would be arbitrary to construe sentences like (52) and
(53) as ungrammatical if there were no motivation independent of sav-
ing one's theory (the standardization thesis) against the opposition (the
ambiguity thesis). It is no argument to say that since the "please"
phenomena can be fit into grammar only if the ambiguity thesis is true,
the "please" phenomena cannot be fit into grammar. Instead of begging
Issues 200
the question, we wish to show how difficult it is to fit the "please"
phenomena into grammar, because of the problem of describing these
phenomena in grammatical terms.
The occurrence of "please" is unproblematic in imperative forms
such as (56) and (57).
(56) Please pass the salt.
(57) Pass the salt, please.
For these cases the grammar can simply restrict the occurrence of
"please" to imperatival contexts. Sentences like (58) and (59) might
seem odd, but they are grammatical:
(58) Please shut up.
(59) Clean the latrine, please.
They are odd because of our attitudes toward what is requested or how
it is requested (compare "He's such a nice little boy" with "He's such a
nice little bastard" -both grammatical), but there is no reason to deny
that they are grammatical.
Problems arise when we try to give a grammatical specification of
when "please" can occur in interrogative forms, for it appears that
reference to the speaker's intentions is necessary, and it is anything but
clear how a grammar can represent information of this sort. To be sure,
the ambiguity thesis provides for a reading of certain interrogative
sentences as whimperatives (as of underlying imperatival form), so it
could be said that no modification need be made in the restriction of
"please" to imperatival contexts. An independent story might be
needed of just which superficially interrogative sentences are the ones
that have this underlying form and can therefore take "please," but
assuming that can be given, there is no special problem about the
occurrence of "please." However, there is another problem. Sentences
like (60) and (61) can be used as requests,
(60) Can you reach the salt?
(61) Will you pass a mailbox?
but if "please" is inserted, as in (60') and (61'),
(60') Can you please reach the salt?
(61') Will you please pass a mailbox?
the ambiguity thesis and its account of "please" in grammatical terms
would require that (60') be taken as an imperative used to request the
Indirect Acts 201
hearer to reach the salt, not pass it, and (61') as an imperative used to
request the hearer to pass a mailbox not to post the speaker's mail. Of
course, a proponent of the ambiguity thesis would say that as requests
to pass the salt and to post one's mail (60') and (61') are not grammati-
cal at all, but then the problem is how the information that the sen-
tences are so used is to be represented in the grammar. A further
problem for the ambiguity thesis is how to block the interpretations of
(60') and (61') as imperatives used respectively to request the hearer to
reach the salt and to pass a mailbox. 'I
"Please" can occur preverbally or postsententially in sentences like II
(62) and postsententially in sentences like (63).
(62) I'd like you to please pass the mustard (please).
(63) I'd like some mustard, please.
It would be preposterous to claim that sentences like these, when used
to make requests, have an underlying imperative form. After all, when
used to make requests, they do so by way of making statements about
the speaker's preference. Whatever plausibility there is to the claim
that whimperatives are not used to ask questions but only to make
requests (hence their imperative reading) is totally lacking in the claim
that (62) and (63) are only superficially indicative. 26
An interesting fact, even if the ambiguity thesis can't explain it, is
that "please" can occur in these sentences when they are used to make
requests, even thoughit couldn't occur if they were used merely to
make statements. A diehard ambiguity theorist might propose that they
have two underlying forms, one merely declarative and one both de-
clarative and imperative. But even if this claim could be made intelligi-
ble, there would still be the problem of accounting for (63), in which the
hearer's requested action is not even mentioned. Perhaps this problem
could be solved by proposing that the underlying form of (63) contains a
verb designating that action ("giving," "passing," "spreading," "cut-
ting"?).
To deal with the "please" phenomena, such as the occurrence of
"please" in (52), (53), (62), and (63), the standardization thesis must
recognize that "please" cannot occur unless the sentence is used to
make a request. If without "please" these sentences are used merely to
ask questions or to make statements, the insertion of "please" makes
no sense and seems to us just plain ungrammatical. Since we claim that
when used indirectly, these sentence forms are also being used literally,
we are committed to the view that with "please" they are ungrammati-
Issues 202

cal, though perfectly acceptable. We construe "please" in these con- Chapter Ten Standardization and
texts (unlike in superficial imperatives) as an illocutionary modifier and Illocutionary Devices
claim that it occurs paragrammatically, as do illocutionary modifiers in
general. That is, "please" modifies what the speaker is doing (indi-
rectly) not what he is saying. 27
We do not yet have an explanation of just when "please" can OCcur
and when it cannot. We could indulge in a bit of hand-waving and say
that since certain forms have become standardly used nonliterally as
requests, they have come to be able to take "please" just as if they
could be used to make literal requests. This is hardly an explanation.
However, we do believe that "please" phenomena, just like other phe-
nomena of illocutionary modification, exemplify a special kind of lin-
What linguistic devices have standardized uses for indirectly perform-
guistic fact, which we dub paragrammatical only to indicate that it
ing illocutionary acts? And what connections are there between par-
cannot be accommodated in grammars as we know them. Along with
ticular linguistic devices and the various illocutionary acts they are
special cases of illocutionary standardization, illocutionary modifiers
used to perform? We begin with the case of simple performatives,
are one of the topics of the next chapter.
perhaps the clearest case in which linguistic material and illocutionary
force are intimately connected.!

10.1. PERFORMATIVES AS CONSTATIVES

I

,I
Austin held (1962, 5) that performative utterances "do not 'describe' or
'report' or constate anything at all, are not 'true or false.''' Rather, they
are, or are part of, the doing of an action. Because the sentence uttered
in a performative utterance is grammatically declarative, Austin's
doctrine once seemed paradoxical. It has lost its air of paradox as phi-
losophers have become more cognizant of the distinction between
sentences and statements and realize, thanks in part to Austin, that not
all sentences are used to make statements. Nevertheless, we wish to
argue that the negative side of Austin's doctrine-that performative
utterances do not constate, are not true or false-is mistaken. Since we
accept the positive side-that they are, or are part of, the doing of an
action-our position is that performative utterances (other than con-
ventionalized ones) are both doings and statings. Thus, they comprise
two illocutionary acts. To utter a performative sentence is to do what
one is stating one is doing; indeed that is what makes the statement
true.
An (explicit) performative is the utterance of a sentence with main
verb in the first-person singular, simple present indicative active,2 this
verb being the name of the kind of illocutionary act one would ordinar-
Issues 204
ily be performing in uttering that sentence (call such a verb a performa-
tive verb). For example, typical utterances of " I order you to leave," "I
promise you ajob," and "I apologize for the delay" are order, promise,
and apology, respectively. 3 Such utterances appear to be of a form
which, with nonperformative verbs, can be used to make true or false
statements, statements to the effect that the speaker is in the state
named by the verb, such as "I see the light" or "I hate spinach."4
Indeed, the use of a sentence with a performative verb not in the
first-person singular, simple present indicative active would ordinarily
be to make a true or false statement: "I ordered him to leave," "(By
signing this) I am promising you a job," and "He apologizes for the
delay." Austin (1962, 63) hopes vainly that this asymmetry will distin-
guish performative from other verbs.
Austin held that despite their declarative grammatical form, per-
formative utterances are not statements,S are not true or false. Rather,
the job of the performative formula is that of "making explicit (which is
not the same thing as stating or describing) what precise action it is that
is being performed by the issuing of the utterance" (1962, 61). And to
use that formula is to perform an act of the sort named by the perform-
ative verb. This seems to be Austin's reason for thinking that per-
formatives (not counting explicit constatives like "I state") are not
constative. That is, in uttering a performative sentence, performing an
act of the sort named by the verb is incompatible with one's also stating
thereby that one is performing such an act. But why cannot one both
perform an act and in the same breath state that one is performing
it? Why should the use of certain verbs in a performative utterance be
any less a statement than the use of the same verbs in nonperformative
utterances, just because this use is also something other than a state-
ment? These rhetorical questions require an explanation of how it is
possible to do both, but first let us examine several arguments that
doing both is impossible.
The following three arguments, which seem to reflect Austin's think-
ing on the matter (he gave no explicit argument), are totally inconclu-
sive because all they show is that a performative utterance is not
merely a statement, not that it is not a statement at all. They seem to
assume that an order, for example, is not a statement just because it is
also something else.
First, (nonconstative) performative utterances are neither true nor
false. Therefore, they are not statements. This argument (Austin 1962,
Standardization 205
12fI) is clearly question-begging. As orders per se (or as promises,
apologies, and so on) performative utterances are neither true nor false.
But if they are also statements, then as statements they are true or
false. Indeed, if true, they are true in virtue of being made.
Second, someone who utters to A "I order you to leave" would not be
said to have stated that he was ordering A to leave. Even if he would not
be said to have stated that he was ordering . . . ,it does not follow that
he did not state that he was ordering A to leave. In fact, he would not
be said merely to have stated that he was ordering A to leave. And that
he would be said to have ordered A to leave does not imply that he was
not stating that he was ordering A to leave.
Third, someone who utters "I order you to leave" does not intend to
convey information, namely, that he is ordering A to leave; he intends to
be thereby ordering A to leave. He could very well be intending and I
doing both. Indeed, we suggest that he succeeds in ordering A to leave
precisely by virtue of stating that he is ordering A to leave. To be sure,
I
conveying this information was not his primary intention, but insofar as
it was necessary to the fulfillment of his primary intention, it too was
intended.
A much subtler argument is offered by Schiffer (1972, 104-110) to
show that performatives are not used constatively. He holds that a verb
used performatively has the same meaning as when used merely de-
scriptively, and that explicit performatives are constative, as he puts it,
in their logical form or conventional force. However, he argues in-
geniously, they are uttered with "something slightly less than their full
conventional force" (p. 109). It is not clear just what this means, but the
argument for it is clear enough. Schiffer shares with Austin the view
that the performative formula makes explicit the full illocutionary force
of one's utterance. If this force includes being constative-what Schif-
fer calls the "full conventional force" -an infinite regress is supposed
to result. Take the utterance, "I order you to leave." If its full illocu-
tionary force includes being a statement, its full illocutionary force is
not being made explicit-only its being an order is made explicit. The
speaker could make explicit the fact that he is making a statement by
uttering "I state that I order you to leave." But on the view that this
utterance was, like the previous one, made with its full conventional
force, he would have been stating that he stated that he ordered. Thus,
he would have still not made explicit the full illocutionary force of his
utterance. Further attempts to make explicit the full illocutionary force
Issues 206
would always leave more to be made explicit, ad infinitum.6 From this
Schiffer concludes that if the full illocutionary force of a performative
includes being constative, its full force cannot be made explicit. But
since (he assumes) the performative formula does make the full force
explicit, it follows that the full force does not include being constative.
On the other hand, linguistic considerations-essentially that a per-
formative sentence has no special grammatical feature and that the
performative verb (or the sentence as a whole) has no special mean-
ing-indicate that being constative is part of the full conventional force
of performatives. Hence a performative is uttered "with something
slightly less than its full conventional force."
The trouble with Schiffer's argument? is his acceptance of Austin's
view that performatives make explicit the full illocutionary force of the
utterance. If the utterance is both an order and a statement, then its full
force is not made explicit by "I order." But that it is an order is made
explicit, and that, of course, is the point of using the performative
formula.
On a related point Austin held, without argument, that although
using the performative formula makes explicit the precise action per-
formed by the utterance, making it explicit is not to state what it is or to
describe it. Granted, there are many ways to make things explicit other
than to state what they are or to describe them. For example, I can
make explicit my gratitude to someone by praising him ("You're so
kind"), by saying "You didn't have to do that," or by returning the
favor. 8 But why isn't the use of the performative formula a statement of
what I am doing, as when I say "I thank you"? After all, in general it is
possible to perform several actions in one fell swoop, so why should the
utterance of "I thank you" not be both a giving of thanks and a stating
that I am so doing? To be sure, the first is my primary intention, but its
fulfillment is abetted by my secondary intention of making the first
explicit.
We hold that the efficacy of the performative formula is not a con-
sequence of the meaning of the performative verbs. Imagine a state of
affairs in which speakers of English did not make performative utter-
ances in using sentences like "I order you ... " or "I warn you ... "
but used them merely to make statements. If such statements could be
made truly and sincerely, something other than the utterance itself
would have to constitute the order or the warning. This could be the
utterance of another sentence or some gesture from a "vocabulary" of
Standardization 207
performative gestures. Then the statement "I order ... " would be
regarded as false if not accompanied by the utterance of an appropriate
sentence or by the appropriate gesture. This method might be ineffi-
cient compared to ours, but there is no reason (except on a hard
meaning-is-use line) to hold that for these speakers words like "order"
and "warn" would differ in meaning from what they mean for us. Mter
all, they call orders or warnings the same things we do, except for the
utterances of sentences that for us, but not for them, are performatives.
Surely for us "order" in "I order" means the same as it does in other
constructions. 9 Moreover, since they use the same sentences we do but
without performative effect, no special grammatical feature can ac-
count for that effect..
As a matter of fact, we have a device other than the performative to
make illocutionary force explicit. An explication can follow the utter-
ance: "Leave; and that's an order." "I will come; and that's a prom-
ise." Here the acts of doing and stating are kept distinct. A speaker
issues an order, and then states that he has done so, thereby making
explicit what he has done. Using the performative formula is to do both
at once, we suggest, and that we have this convenient formula at our
disposal is not due to the meaning of the performative verbs.
Why then is "I order you to leave" an order, whereas "It is the case
that I order you to leave" and "I state that I order you to leave" are not
orders? The latter two utterances are mere statements, true or false
depending on whether or not the speaker gives the addressee an order
at approximately the time of utterance, by performing or having just
performed some other act, verbal or otherwise, such as forcefully
pointing to the door. They can be true only if there is something other
than themselves to be about. Now suppose that an utterance of"I order
you to leave" is a statement. Suppose, moreover, that the speaker is
performing no other act that could even remotely be construed as giv-
ing an order. In this event, either the speaker is mistaken or he must be
doing something that counts as an order, and the only candidate for this
is his utterance. It itself is the only thing for it to be about. That is, as a
statement, the utterance is about itself, as an order. There is nothing
paradoxical about this, despite the utterance's self-reference. Mter all,
its self-reference can be made explicit by using the word "hereby," as
in "I hereby order you to leave." What "hereby" adds to the original is
something like this: "In uttering this sentence, I order you to leave." 10
The speaker thereby makes explicit not only the force of his utterance
Issues 208
(that it's an order) but the vehicle of that force, namely, the utterance
itself. 11
So far we have rejected known reasons for denying that performative
utterances are statements too; we haved denied that the efficacy of
performatives is a matter of meaning; and we have suggested, with the
help of the "hereby" argument, that taking a performative utterance as
a statement may explain why it is a performative as well. We have not
yet explained why an utterance like "I order you to leave" is a per-
formative. To do this is not to explain historically how or why there
came to be the performative practice, but to show what has to be the
case for such an utterance to count as an order.
The communicative presumption among users of a language is that
when they say something, what they are doing in saying it is determin-
able by their audience. Because this is a matter of mutual belief, the
speaker can reasonably intend the hearer to take him as intending his
act to be determinable, and it is on this basis, together with the ut-
terance itself and the circumstances surrounding it, that the hearer
determines what that act is. The speaker succeeds insofar as this de-
termination is made correctly. In the case of performative utterances,
even those without the use of "hereby," normally the hearer could
reason, and could be intended to reason, as follows:
1. He is saying "I order you to leave."
2. He is stating that he is ordering me to leave.
3. If his statement is true, then he must be ordering me to leave.
4. If he is ordering me to leave, it must be his utterance that constitutes
the order. (What else could it be?)
5. Presumably, he is speaking the truth.
6. Therefore, in stating that he is ordering me to leave he is ordering me
to leave.
This reasoning is artificially elaborate-or needlessly explicit-but
that is only because there is ample precedent for it. The performative
practice short-circuits the steps of this inference pattern, both as in-
tended by the speaker and as carried through by the hearer. Still, the
success of the performative would be vitiated if any of the steps in the
inference were blocked. Explicit performative utterances are indirect
illocutionary acts. 12 The hearer's intended inference, warranted by the
communicative presumption, is compressed by precedent. The explicit
performative formula is standardized for the indirect performance of
the illocutionary act named by the performative verb.
Standardization 209
10.1.1. Locutionary Performatives
It has not been previously observed that there is a class ofperformative
verbs that are not illocutionary but locutionary. These verbs of saying
are performative in that sentences in which they occur in the first-
person present can be true of their very utterance, as in the case of (1)
and (2),
(1) I repeat that there will be no fooling around.
(2) I close by saying that, hard as it is, this problem can be solved.
Among the locutionary performatives are "say," "utter," "comment,"
"mention," "note," "remark," "repeat," "add," "begin," "continue,"
"close," "list," "enumerate," "formulate," and "reformulate." Locu-
tionary performative utterances are true just in case what the speaker
says in issuing them is what he predicates of himself. Unlike communi-
cative illocutionary performatives, no R-intention is required of the
speaker nor need be recognized by the hearer. Unlike conventional
illocutionary acts no special nonlinguistic convention is involved. That
the speaker is mentioning something, repeating himself, or listing items
depends merely on what he says and is identifiable thereby. Of course,
the speaker will also be performing, in issuing a locutionary performa-
tive, some illocutionary act, and this will fit the SAS like any other. We
will say nothing further about locutionary performatives, but it is well
to note that not every performative verb is an illocutionary verb.
10.2. EMBEDDED AND HEDGED PERFORMATIVES
In his reply to Ross (1970), Fraser (1971, 2) noted apparent counter-
examples to the claim that the performative verb must be the highest
verb in surface structure:
(3) I regret that I must inform you of your dismissal.
(4) I am pleased to be able to offer you the job.
(5) I would like to congratulate you.
Since on our view performativity is indirect even in explicit performa-
tive utterances, these sorts of sentences offer no special problems for
our account. Embedding increases the inferential load on the hearer,
but there is no difference in kind between performative utterances with
unembedded and those with embedded performative verbs. Sadock
(1974, 55-61) has in effect argued against the indirectness of these
Issues 210
cases, and Fraser (1975) has gone on to investigate sentences like these
in some detail under the label of hedged performatives.
10.2.1. Embedded Performatives
Sadock (pp. 56-61) gives four arguments that are supposed to cast
doubt on the view that illocutionary acts performed by uttering sen-
tences like (3) -(5) are performed indirectly.
First, suppose that Mr. Frambes has received a letter from his insur-
ance company beginning,
(6) a. We regret to inform you that your policy is canceled.
After having an accident he takes the company to court, claiming that
the company simply expressed a certain negative psychological state-
regret at informing. Sadock concludes "Who wins? Pretty obviously,
the insurance company does" (p. 56). Maybe so, but why does this
count against the indirect hypothesis? Surely the law does not require
that the policy holder be informed by a direct rather than indirect
illocutionary act.
Second, Sadock notes that one can report the act performed using
(6a) with sentences like (6b), but sentences like (6c), (6d), and (6e) are
misleading, incomplete, or ungrammatical:
(6) b. Mutual of Hoboken informed Mr. Frambes of the cancellation of
his policy.
c. Mutual of Hoboken canceled Frambes's policy.
d. Mutual of Hoboken expressed their regret about informing
Frambes of the cancellation.
e. *Mutual of Hoboken regretted to inform Frambes of the cancel-
lation of his policy.
Sadock concludes that "It would seem that the underlying clause
whose main verb is inform is, after all, a performative clause" (p. 57).
And if this is right, the act would be for Sadock literal and direct, not
indirect. There are some objections to this argument. There is a gram-
matical version of (6e):
(6) e'. Mutual of Hoboken regretted informing Frambes of the cancel-
lation of his policy.
More seriously, Sadock's reasoning would rule out the report of indi-
rect illocutionary acts by the use of any performative verb. But surely
Standardization 211
one can report that someone requested a drink in uttering "My mouth
is parched," although there is no underlying imperative or verb of
requesting.
Third, Sadock notes that (6a) will take "hereby" as in (6f):
(6) f. We regret to inform you that your policy is hereby canceled.
Apparently this is supposed to count for the directness and literality of
the act, but Sadock does not say why, and we have already offered a
different account of "hereby" (see section 10.1).
Fourth and finally, Sadock claims that verbs like "regret" "require
that their complements have all the salient properties of explicit per-
formatives" (p. 59). For us, of course, this may well be true and still not
preclude the possibility of indirection in the utterance of (6a). Again,
we are given no argument to the contrary.
We conclude that these arguments cast little if any doubt on the view
that such acts as illustrated above can be indirect.
10.2.2. Hedged Performatives
Fraser (1975) has discussed the interesting case of utterances that differ
from simple performative utterances in that the performative verb is
preceded by a modal like "must," "can," "will," "would," "might,"
"should," or a semimodal such as "have to." Such utterances seem to
have the illocutionary force of the act named by the performative verb
used, as illustrated by typical utterances of sentences like (7) - (1 0),
which Fraser calls hedged performatives:
(7) I must ask you to leave.
(8) I can promise you I'll be home.
(9) I want to thank you for the Beaujolais.
(10) I would suggest you try some.
Fraser is concerned both to account for their illocutionary force in
terms of certain conversational principles, and to account for cases
("weak performatives") that do not have the force of the act named by
the performative verb, such as "I must forbid you from cutting off your
right arm." Fraser seems to assume (he says nothing explicitly) that
simple performatives do not pose the same explanatory problem as
hedged performatives. Since he indicates nothing to the contrary, pre-
sumably he thinks that simple performatives are literal and direct illo-
cutionary acts and therefore that they have their illocutionary force
Issues 212
solely in virtue of literal sentence meaning.13 We have argued that
simple performatives are indirect and that literally they are constatives;
we take the same position on hedged performatives. Nevertheless,
hedged performatives do contain modals, whose distinctive functions
need to be explained.
Consider Fraser's approach to cases involving the modal "must." He
proceeds on the assumption that utterances of sentences like (7) are
statements that the speaker has a certain obligation (he does not men-
tion that it may be otherwise necessary that he do something)-in this
example, to ask the hearer to leave. The question is why such an
utterance counts as the performance of the act the speaker says he is
obligated to do. Fraser (1975) posits three principles:
Principle of Obligation Fulfillment. Given nothing to suggest the con-
trary, whenever someone has an obligation to perform some action,
one can infer that he will perform that action.
Principle of Unspecified Time. Given nothing to suggest the contrary,
whenever the time of an action is left unspecified, one can infer that the
,agent is expected to perform the action at the earliest chance.
Principle of Efficiency. Given nothing to suggest the contrary, when-
ever a further utterance would be redundant, one can infer that the
speaker need not make the utterance but that he will operate as if he
had made it and will expect the hearer to operate similarly.
Fraser does not discuss the epistemological or conversational status of
these principles. To sanction the intended inference, though, they need
to be mutual beliefs between speaker and hearer.
The first principle (obligation fulfillment) is weak by itself, but in
conjunction with the second (unspecified time) it may be inferred that
the speaker of (7) will, at the earliest chance, ask the hearer to leave.
The third principle (efficiency) is needed to account for the fact that the
utterance of (7) is in itself the fulfillment of the obligation that the
speaker is ascribing to himself. However, the formulation of this prin-
ciple makes it unclear how and why the utterance of (7) constitutes an
act of asking the hearer to leave. If anything, it suggests that no such
act need be performed once the effect of that would-be act has been
achieved. However, matters are made worse by the use of the word
"redundant." It is redundant for a speaker who says "I must ask you to
leave" to ask the hearer to leave only if he has indeed asked the hearer
to leave, or if his utterance at least has the effect (expresses the illocu-
tionary intent) of asking the hearer to leave. Perhaps these problems
with the principle of efficiency can be removed if we follow the spirit of
Standardization 213
one of Grice's (1975) maxims of manner, "Be brief." Grice means
brevity in saying what one has to say, but allowance could also be made
for illocutionary brevity. A maxim to this effect would take the form,
"Don't go on to perform an illocutionary act directly when its intent
can be inferred from the utterance you have just (or already) made." If
such a maxim applies to ordinary conversational situations, then the
speaker of (7) need not explicitly request the hearer to leave, provided
the intent to make such a request can be inferred from the utterance of
(7). Obviously our maxim of illocutionary brevity tells us nothing about
when indirect illocutionary intents can be so inferred.
Recall that one of the basic ideas underlying the speech act schema is
the implication of the communicative presumption that there is a rec-
ognizable sufficient reason (explanation) for the speaker's utterance. If
there is no apparent sufficient reason for an utterance to be taken
literally, the hearer must search for a nonliteral or indirect interpreta-
tion of the utterance for which there is sufficient reason of utterance.
The operant principle would take the form "Don't issue an utterance
for which (taken literally) there is insufficient recognizable reason un-
less you have some further (or other) illocutionary intent that can be
inferred." This still does not tell us how the inference is to be drawn,
but surely it must be based, as the SAS requires, on the content of the
utterance and on the mutual contextual beliefs that obtain between the
speaker and the hearer. How would the inference run in the case of (7)?
As usual, the inference is one that S intends H to make partly on the
basis of recognizing S's intention that he make it. For the sake of
simplicity, we will omit the steps preceding the literal illocutionary act.
1. S is stating that he must ask me to leave.
2. S's stating that he must ask implies that he is reluctant 14 to do so.
3. S's having to ask me to leave is a reason for so doing.
4. S's stating that he has this reason is reason to think that he intends to
ask me to leave.
5. S is not explicitly asking me to leave.
6. In stating that he must ask me to leave, S intends thereby to be
asking me to leave.
7. S is asking me to leave.
Why do only certain utterances of the form "I must V ... ," where
V is a performative verb, count as acts of V -ing? Fraser argues quite
plausibly that when the speaker does not want to avoid responsibility
for V-ing (because he is not reluctant) as with typical utterances of (11)
 Issues 214
and (12), the utterance is not clearly an act of V-ing (Fraser calls it
"weakly performative"):
(11) I must congratulate you on winning.
(12) I must welcome you home.
There is no point in stating that you must do what you don't mind
doing, hence no reason to use the hedged performative with "must."
Indeed, not only is there no reason, the utterance will not be taken as
an act ofV-ing if the hearer does not think V-ing is a sort of act S should
be reluctant to do. Thus, if you arrive home and you are greeted with
an utterance of (12), you are likely to expect either a reluctant wel-
coming sequel or no welcome at all. If anything, then, when "must"
precedes the performative verb, the speaker can be taken as expressing
his reluctance to V. The utterance counts as an act of V-ing that p only
if it is the sort of act that people in general are reluctant to do or if it is
mutually believed between Sand H that S is reluctant to do it. Notice
that the question of the speaker's reluctance figures in the inference H
is intended to make as to S's illocutionary intent. For this reason, the
inference is blocked if H has reason to believe, and to believe that S
believes he believes, that S is not reluctant to V.
We will not review Fraser's account of the other main kinds of
hedged performatives. In each case he posits certain conversational
principles that are supposed to account for the force of the utterance.
Unfortunately, he fails to justify these principles, although he repre-
sents many of them as being akin to Grice's principles. That of course
does not mean they are, and even if they are, that does not make them
true. He does not discuss their conversational or epistemological sta-
tus. Some are formulated as inference principles of the form "Given
such-and-such, one can infer so-and-so." But what validates such prin-
ciples? What makes them applicable to particular conversational situa-
tions? On our view they have to be mutually believed to be applicable;
otherwise they could not figure in intended hearer inferences. Since
Fraser himself disavows (1975, note 6) any claim to finality in his for-
mulation, we will not quibble about details. Rather, we have tried to
show in the case of "must" that there is no need to appeal, either as
theorists or as speaker-hearers, to such principles. The communicative
presumption is quite sufficient to account for hedged performatives in
that it requires the hearer to search for a sufficient reason for the
speaker's utterance. Just as in the case of "must," the other kinds of
Standardization 215
hedged performatives can be understood in terms of recurrent patterns
of inference to sufficient reasons for utterance.
Therefore we might suggest a general maxim to cover all the kinds of
hedged performatives, whose exploitation follows Grice's pattern of
conversational implicature:
Maxim of Sufficient Reason: Make your utterance such that there is
identifiably sufficient reason for its issuance.
Grice's maxims do not govern illocutionary acts generally, only acts of
constating. He intimates that they can be generalized and we suggest
the maxim of sufficient reason as one such. Indeed it is central because
in effect it requires the speaker to utilize the communicative presump-
tion in making his utterance. In the case of hedged performatives, and
for indirect illocutionary acts generally, this maxim is utilized in just
the way that Grice's are for conversational implicature: the speaker
intends his utterance to be taken literally as a certain illocutionary act,
identifiable on the basis of what is said, and it will be viewed by the
hearer as being issued with sufficient reason only on the assumption
that there is some other illocutionary act being performed in the pro-
cess. That is just what happens in the case of hedged performatives.
For hedged performatives, like standardized indirect acts generally,
there is ample precedent for the inference the hearer is intended to
make; consequently the SAS is short-circuited. What distinguishes di-
rect performatives and hedged performatives from illocutionary stan-
dardization generally is that the illocutionary verb explicitly occurs in
the utterance. Thus the hearer's search procedure, even if such utter-
ances lacked precedent, would be simple and short. In reviewing the
other main cases of hedged performatives, we will give brief versions of
the inference required in accordance with the maxim of sufficient
reason.
According to Fraser, hedged performatives with "can" generally re-
quire some adverbial like "now," "finally," or "at last" to count as the
sort of illocutionary act named by the performative verb:
(13) I can now admit that I did it.
Such an adverbial is not necessary, however, as shown by example (8),
repeated here:
(8) I can promise you I'll be home.
Issues 216
Whether or not the sentence includes temporal specifiers like "now," it
must have sufficient reason for being uttered. We agree with Fraser
(except that he needlessly posits a "principle of expressed ability" to
account for it) that the hearer is to think, "The speaker wouldn't have
told me this unless he intended to do it." Instead of invoking Fraser's
principle of efficiency to account for the fact that the utterance itself is
the execution of that intention, we can say that the only identifiable
sufficient reason for the utterance is that it be the execution of that
intention, as indeed a hearer would recognize by reasoning (invoking
the maxim of sufficient reason) roughly as follows: 15
(13) a. He is stating that he can now admit that he did it.
b. He wouldn't so state unless he intended to.
c. If he intends to admit that he did it, he will.
d. No (additional) utterance to that effect is forthcoming.
e. Therefore, in stating that he can, he is admitting that he did it.
One question not to be overlooked is why, at least in some instances,
replacement of the modal with a periphrastic version takes away the
indirect illocutionary effect. Compare the following two sentences:
(14) I can promise you I won't squeal.
(15) I am able to promise you I won't squeal.
An utterance of (15), unlike (14), would not normally be taken as a
(hedged) promise but simply as an assertion of the speaker's ability.
Not only that, it would imply (at least if so taken), contrary to the
analog of (13b), that the speaker does not intend to make such a prom-
ise. However, the same effect is achieved by an utterance of (14) in
which "can" is stressed:
(14') I can promise you I won't squeal (but that doesn't mean I will
promise).
Fraser rightly points out that both the stressed and the periphrastic
version call attention to the assertion of ability (the direct illocutionary
act), but he mistakenly concludes that they suggest that there is some
reason to doubt the speaker's ability. Rather, they implicate that S is
not willing to, and is not going to, do what he says he can do. 16
The pattern of inference suggested for "can" cases is applicable to
cases of "will"/" shall"/" am going to"/"intend to," except that step (b),
where the speaker's intention is inferred, is omitted since what the
Standardization 217
speaker is stating is that he intends to do something. Otherwise, these
cases require no special comment: if no further utterance is forth-
coming, the speaker is doing what ,he states he intends to do. Notice
that adverbs like "now" and "hereby" can be inserted in such sen-
tences as (16), just as with simple performatives.
(16) I will {now } propose going home.
hereby
,I
I
We are not suggesting that "will" in this context is a present auxiliary
I
I
verb rather than a future one, but we do suggest that the speaker's
reference to the time of his forthcoming proposal does not extend be-
yond the time of completion of his utterance.
Cases involving "want to" /"wish to" /"would like to" have a rather
special feature. Fraser points out that sentences like (17) are taleen as
requests for permission, or, we may add, for cooperation.
(17) I want to ask you a question.
Since utterances of such sentences are literally statements of what
the speaker wants, they are indirect requests. Therefore, utterances
like (18) would be doubly indirect.
(18) I want to ask you if you've seen Jaws.
For Fraser such an utterance is literally a statement, indirectly a re-
quest (for permission), and indirectly a question. Now Fraser posits a
principle of permission seeking to account for the inference from the
request to the performance of the requested act, in this case a question.
Aside from our usual problems with his principles, we suggest that the
indirect question works directly off the literal statement of want, not off
the indirect request. Thus, the pattern of inference is the same as with
"will" and other statements of intention. Indeed it seems that no indi-
rect request at all is being made, because the utterance has the force of
a question just because it is presumed by the speaker that no permission
is necessary for such a question.
The hedged performatives with "would"/"might"/"should" are in-
teresting because grammatically they seem to be consequents of sub-
junctive conditionals without any expressed antecedent. A -sentence
like (19) might be constIued as an elliptical version of a conditional,
with an antecedent like "If you were to ask my opinion."
(19) I would suggest a shot of Irish whiskey.

Issues 218
Fraser claims further that there is no need for any conversational prin-
ciples to interpret such examples and that they can be interpreted as if
the "would" were absent: according to Fraser, the illocutionary force
(in the case of (19), as a suggestion) cannot be denied by the speaker.
Unfortunately, he does not state whether the result of denial is a con-
tradiction or an ungrammaticality. We are not sure what to claim here.
Consider the cancellation of (19) as in (19').
(19') I would suggest a shot of Irish whiskey, but I won't.
Without the antecedent made explicit, it is not clear what the point of
an utterance of (19') could be. (For that matter, it is not clear what the
point would be with an explicit antecedent but without an explana-
tion-say, that the speaker is out of Irish whiskey.) In any case, we see
an inadequacy with this account as it stands. What grammatical reason
is there that the suppressed antecedent must be something like "If you
wanted my opinion"? If the suppressed antecedent were "If I approved
of alcohol," the utterance of (19) would not count as a suggestion (to
someone suffering from a sore throat), but this is not a grammatical
fact. So it appears that reference to conversational principles is needed
here after all.
At the end of his paper, Fraser mentions but does not discuss cases
like (20) and (21) and doubly hedged performatives like (22) and (23).
(20) I am happy to inform you that you're accepted.
(21) I hasten to add that I didn't see her either.
(22) I will have to request that you leave at once.
(23) I should now be able to assure you that this will work.
He suggests that further conversational principles are needed for these.
Our position, as should be clear, is that no special conversational prin-
ciples are needed to account for the performativity of any of these
cases. Indeed hedged performatives are not fundamentally different
from simple performatives. In both cases the hearer merely has to
identify a sufficient reason for the speaker's utterance. Considering the
occurrence in the sentence uttered of the verb designating the very type
of act being performed (as well as precedent for such performances-
see 9.3) it is no surprise that the identification can be made.
FinaJly, Fraser claims that for each kind of hedged performative only
certain sorts of illocutionary verbs work performatively with the modal
(or other expression) in question. The following, for example, are
clearly not acts of the sort named by the performative verb.
Standardization 219
(24) I must invite you to stay.
(25) I can (now) ask you to go.
(26) I will order you to sit down.
However, it is not the verb itself but the verb together with its comple-
ment that determines whether an utterance of a sentence in normal
circumstances has the illocutionary force designated by the verb. The
following examples, with the same verbs as in (24)-(26), seem to have
that illocutionary force:
(27) I must invite you to another one of those horrible parties at the
boss's house.
(28) I can (now) ask you to turn up the radio-the kids are asleep
finally.
(29) I will order you never to come back until I tell you to.
In each case the hedged form has the illocutionary force named by the
verb because the utterance meets the conditions that hedged performa-
tives of that sort meet. Hedged performatives with "must" are acts that
the speaker is reluctant to do. Those with "can" imply that the condi-
tions were not right previously but are now. Those with "will" imply
that the speaker was not previously willing or that a certain condition is
assumed under which he is willing. And so on for other cases.
The point is that performativity, simple or hedged, is not a question
of semantics. The only question about the acceptability of a performa-
tive utterance is whether there is reason for it-identifiable reason.
Contrary to some of the literature, there is no need to posit such a thing
as a performative reading of sentences used performatively, that is, to
perform an act (indirectly in our view) of the sort named by the per-
formative verb. When such a sentence is so used on a given occasion,
no special reading is required to explain that use. The explanation is
pragmatic, not semantic.
10.3. ILWCUTIONARY ADVERBIALS
Many locutions can be used to comment upon the illocutionary intent
behind the utterance of the very sentences in which they occur. Con-
sider the following:
(30) Frankly, you bore me.
(31) If I may say so, this conversation is getting tedious.
(32) By the way, I couldn't find your underwear.
Issues 220
(33) Speaking of linguistics, did you know that procedural semantics is
a notational variant of Fortran?
(34) If you're so smart, who is the voice of Bugs Bunny?
(35) On the other hand, he who hesitates is lost.
(36) Moreover, we have no bananas.
(37) Since you'll find out anyway, your wife is carrying on with the
butler.
Each of these sentences appears to be a perfectly grammatical and fully
meaningful English sentence, and yet the prefatory adverbial is not
used to modify the main clause of the sentence (it may not even con-
tribute to the locutionary act). Rather it is used to characterize, in one
way or another, the utterance of the main clause. In (30) "frankly"
describes S's act of stating that H bores S. "By the way" in (32)
indicates the digressive character of the utterance to follow. "On the
other hand" in (35) and "moreover" in (36) indicate, respectively, that
what follows contradicts or supplements something said previously. In
(37) the prefatory clause" Since you'll find out anyway" supplies part
of a reason for making the ensuing statement; obviously it provides no
reason for believing what is stated. In the case of (34) the J?refatory
clause provides a reason for the addressee to answer the question that
follows.
There are many types of illocutionary adverbials, as we call them.
What is important about them is not their variety, interesting as it is,
but the issues they raise regarding the relation of linguistic theory to
language use. As illustrated by our discussion of illocutionary stan-
dardization and of performativity, we maintain that there is a place
where linguistics leaves off and the theory of illocutionary acts takes
over. The only concessions we have made to the use-is-meaning view,
shared by many generative semanticists and speech-act semanticists, is
to allow that the type of saying, as determined by the grammatical
sentence type (declarative, imperative, or interrogative) delimits literal
illocutionary force potential and that there is a semantic connection
between the meaning of indexicals and how people use them to refer.
Apart from these concessions, we have enforced a strong use/meaning
dichotomy. However, it seems that the existence of utterance adverb-
ials collapses this dichotomy and that a linguistic account of sentences
involving them must make reference to the use of those sentences.
The higher performative analysis is an approach to the problem of
illocutionary adverbials which identifies use and meaning. For exam-
ple, on this theory (31) becomes (31P), with embedded performative.
Standardization 221
(31P) If I may say so, I say that this conversation is getting tedious.
And (34) becomes (34P):
(34P) If you're so smart, I request that you tell me who is the voice of
Bugs Bunny?
In both cases the performative analysis provides a syntactic place for
the if-clause, in which the semantic role of that clause is straight-
forward. Can this sort of analysis be motivated for the rest of (38) -(45)
We will look briefly at some representative suggestions by Schreiber
(1972) and Sadock (1974) concerning such cases. First we will compare
some manner adverbs ("frankly," "truthfully") with some sentence
adverbs ("fortunately," "clearly") and argue that their distribution
does not support a performative explanation (though it may be con-
sistent with one). Instead we offer an illocutionary explanation of the
data. We then generalize our position to expositive and reason adverb-
ials such as "If I may say so" and "Speaking of Jones."
10.3.1. Manner Adverbs
Consider Schreiber's (1972) application of the performative analysis to
manner adverbs, a special case ofillocutionary adverbials illustrated by
(30) and by (38) and (39):
(38) Truthfully, you lied to me.
(39) Truthfully, did you lie to me?
After arguing that manner adverbs like "truthfully" and "frankly"
differ syntactically from sentence adverbs like "fortunately" and
"clearly" (clauses in which the former occur, unlike those in which the
latter occur, cannot be embedded in predicate complement construc-
tions!7), Schreiber supports the performative analysis by suggesting
that it accounts for the fact that in (38) the speaker is predicating
truthfulness of himself, whereas in (39) truthfulness is predicated of the
addressee. Indeed on the performative analysis this asymmetry is ex-
posed neatly when (38) and (39) are derived from, respectively, (38P)
and (39P):
(38P) I tell you truthfully that you lied to me.
(39P) I request that you tell me truthfully whether you lied to me.!S
It is evident that in each case a syntactic place is provided for "truth-
fully" to modify just what it is supposed to modify.
Issues 222 Standardization 223

Greenbaum (1969,84) noted that sentence adverbs like "fortunately" S

and "clearly" do not occur freely in sentence-initial position in ques-
tions.
(40) *Fortunately, did you lie to me?
(41) *Clearly, did you lie to me? +perf ]
[ + order - - - - V ------=:::::..
Sadock (1974, 34) has converted this sort of observation into the fol-
lowing argument for the presence of an abstract (higher) performative [+ adverb]
verb:
1. Sentence adverbs (such as "fortunately") do not occur with impera-
tive sentences: *"Fortunately, leave the room!"
2. Sentence adverbs "begin life as the subject clauses of predicates that Figure 10.1
:,I
express adverbial ideas."
3. A verb of ordering demands that "its indirect object and the subject s
of its complement clause be coreferent[ial]."
4. Conclusion: "An imperative with a sentence adverb would have to
stem from a structure such as [figure 10.1] which does not meet the
condition on semantic well-formedness, since NPi, which refers to an [+adverb]
individual, cannot be coreferent[ial] with NPk, which is a proposition." s
This argument cannot be accepted as it stands because the generaliza-
tion in 3 mentions indirect object position (NPv), while the conclusion
mentions subject position (NPi), a problem that can be easily remedied
by changing NPi to NPv in the conclusion on the plausible assumption
Figure 10.2
that NPv too must refer to a person or at least a nonproposition. But
even with this modification the argument does not motivate the conclu-
sion, since no argument is given for why the adverb cannot come from proposal would have it, "truthfully" modifies the deep performative
a higher predicate, as in figure 10.2. As Sadock says, "it has been clause. 19 But why does "truthfully" not modify "you lied to me"? Not
supposed that adverbs in general are reduced versions of higher only does it obviously modify the main clause in (44) and (45)-
clauses" (p. 34). (44) Truthfully, you answered the questlon.
That the argument for an abstract performative verb is consistent (45) Truthfully, you conveyed how you felt.
with the data doesn't make it sound. Compare (42) and (43) with (38),
"Truthfully, you lied to me." - but the only reason it seems not to modify the main clause in (38) is
that a contradiction would result, as is evident if "truthfully" is post-
(42) Fortunately, you lied to me. posed, as in (38').
(43) Clearly, you lied to me.
(38') You lied to me truthfully.
In these two cases the sentence adverb modifies the sentence "you lied
to me," whereas in (38) "truthfully" modifies, so to speak, the illocu- But surely this does not mean that (38') is not a literal paraphrase of
tionary act of uttering that sentence or, as the abstract performative (38). We see no reason to deny that it is. The question is whether (38)
Issues 224
has two literal readings or only one, the one paraphrased by (38') and
syntactically like (42) and (43). If there are two, then (38) is syntacti-
cally ambiguous. However, this supposition seems to require the higher
performative analysis, and we have just seen its difficulty in handling
the data. Therefore the other option is worth examining. It requires the
assumption that the use of manner adverbs like "truthfully" as illocu-
tionary modifiers is nonliteral. Since this phenomenon is general in
scope, it looks like a case of illocutionary standardization, analogous to
the cases of explicit and of hedged performatives.
A grammatical form is a case of illocutionary standardization if there
is ample precedent for using it to perform an illocutionary act of a
certain type. Illocutionary standardization is not a matter of linguistic
fact, at least in the narrow sense of being established by a linguistic
rule. On our view standardization short-circuits the SAS by by-passing
the steps involved in determining the literal illocutionary act. Indeed, in
some cases we will now consider there is no literal illocutionary act to
be identified, for the sentence uttered makes no sense literally. It is
grammatical, however, from a purely syntactic standpoint, and a stan-
dardized form of inference is R-intended to be made to determine the
speaker's illocutionary intent. In tlie case of (38), "Truthfully, you lied
to me," the sentence uttered makes sense literally, but the presumption
of sincerity invites H to find an alternative interpretation of the utter-
ance, which would be literally rendered as "I tell you truthfully that
you lied to me," the nearest plausible candidate for what the speaker
could have meant. This account is no mere notational variant of the
performative analysis: the locus of explanation is not in the grammar
but in the social psychology of the situation. A theory of linguistic
performance might have to account for the inference made, but this
does not make it a matter of grammar.
Where the presumption of literalness is overridden, the hearer relies
on the presumption of sincerity in searching for a nonliteral intention.
For example, the manner adverbs we have considered cannot be
plausibly construed as modifying the verb of the sentence uttered and
in some cases they cannot even be sensibly so construed. In either
event, the strategy is to take the preposed adverb as modifying the
utterance of the sentence. There is no more a priori reason to assume
this to be impossible than there is to posit gratuitous constructions like
higher performatives. Pretheoretically, it is intuitively plausible to re-
gard these manner adverbs, as well as the other sorts of adverbials in
Standardization 225
examples (31)-(37), as illocutionary modifiers, but the theoretical
question is how to deal with this intuition linguistically. The performa-
tive analysis takes the bull by the horns and posits a linguistic form for
the illocutionary act modified. Our approach, on the other hand, denies
that illocutionary acts so modified are fully literal, much less that they
have linguistic forms. They have literal correlates, the utterance of
precisely those sentences occurring in performative paraphrases, but
no higher performatives are posited in the sentences actually uttered.
Our argument for denying that sentences with preposed manner
adverbs have literal meanings of the sort assumed by the performa-
tive analysis can be summarized as follows. Manner adverbs can be
straightforward sentence adverbs, like "clearly" or "unfortunately."
However, when they function as illocutionary modifiers, they cannot
be taken literally as sentence adverbs because such an interpretation is
nonsensical. Relying on the presumption of sincerity, together with the
overarching communicative presumption, the hearer takes, as he is
intended to take, the uttered sentence as lexical shorthand for what
would be said explicitly using the full performative form. There is no
reason to assume or postulate that that form underlies the sentence
actually uttered.
To deny that sentences like (30) and (38) with manner adverbs con-
tain higher performative verbs requires us to deny that they are gram-
matical. Intuitively, such sentences seem grammatical, and yet we are
denying that they are. Our theoretical justification is that there is noth-
ing in the sentence for the manner adverb to modify. (Of course, a
proponent of the higher performative approach disagrees- but only by
inventing something for the adverb to modify.) Our proposal is indeed
controversial, but as we argue in section lOA, a wide variety of expres-
sions, extending far beyond those even seemingly amenable to the
higher performative approach, can and commonly do serve a conver-
sational purpose without being grammatical. Usability is not gram-
maticality, and acquiring a use does not tum the ill-formed into the
well-formed.
10.3.2. Generalizing the Argument
The pattern of argument for manner adverbs can be generalized to
cover all sorts of illocutionary adverbials, including those illustrated in
examples (31)-(37). If anything, the argument is more compelling in
Issues 226
these cases, because the alternative, the higher performative approach,
requires greater theoretical contortions and is supported by fewer data
than for manner adverbs.
The performative approach is relatively plausible for case (31); at
least the performative paraphrase is easy to formulate.
(31) If I may say so, this conversation is getting tedious.
(31P) If I may say so, I say that this conversation is getting tedious.
Taken literally, (31) contains an inappropriate juxtaposition of ante-
cedent and consequent, although it is not clear that this inappropriate-
ness is linguistic- the conditional just seems absurdly false and the
literal utterance of it pointless. To argue that (31) is derived from (31P)
it must be argued either that this literal interpretation is based on a false
grammatical reading of (31) or that it is based on a genuine but irrele-
vant reading, in which case (31) is held to be syntactically ambiguous.
The first alternative makes the absurdity of the conditional a matter of
grammar when obviously it is not, because a hypothetical case, how-
ever wild, could be constructed in which the conditional is true and its
utterance appropriate: suppose that S is permitted to speak only when
conversational tedium sets in. The latter alternative has the onus of
showing just how and when an interpretation of form (31P) is required.
And if the performative analysis requires higher performatives under
both interpretations, then it has the onus of explaining when sentences
like (31) are to be read as (31P) and when they are to be read as (31P'):
(31P') I say that, if I may say so, this conversation is getting tedious.
This would correspond to what we have been calling the literal in-
terpretation, and if it seems less than acceptable, try the two performa-
tive readings of (46):
(46) If I may continue, this conversation won't get tedious.
(46P) Ifl may continue, I say that this conversation won't get tedious.
(46P') I say that, if I may continue, this conversation won't get tedious.
How, on the performative analysis, can performative deletion be freely
allowed regardless of where the higher performative is located in the
complex sentence?
The adverbial in (32) seems clearly not to be a sentence adverb.
(32) By the way, I couldn't find your underwear.
Standardization 227
However, clearly it is in (32'), which seems to be a more natural para-
phrase of (32) than (32P).
(32') By the way, I might mention that I couldn't find your underwear.
(32P) By the way, I say that I couldn't find your underwear.
On the performative analysis, is "by the way" a sentence adverb only if
an explicit performative occurs in the sentence and otherwise a modi-
fier of the deeper performative? After all, a sentence like (32') is de-
rived, on the performative analysis, from (32'P):
(32'P) By the way, I say that I might mention that I couldn't find your
underwear.
Again it seems that a nonliteral reading of (32), in accordance with an
inference parallel to those required in the previous cases, is sufficient to
account for the illocutionary force of its utterance.
With the possible exception of (37) the remaining examples do not
lend themselves to straightforward performative paraphrases at all.
Consider, for example, (33):
(33) Speaking of linguistics; did you know that ... ?
(33P) Speaking of linguistics, I ask you did you know that ...
The performative paraphrase (33P) really doesn't do justice to (33)
because the dangling phrase "speaking of linguistics" describes not the
speaker's subsequent utterance but the subject matter of the stage of
the conversation at which the utterance occurs. Roughly, the force of
this phrase is to give a hedged assurance that what follows is relevant to
the conversation, that is, it is but it isn't.
We might use the term conversational paraphrase to designate a sen-
tence whose literal utterance has the same force as an utterance of a
sentence like those under consideration. Unlike the performative para-
phrase, which supposedly corresponds to a deeper level of linguistic
representation, conversational paraphrases can take a variety offorms.
The only constraint is that they be readily identifiable so that Scan
reasonably expect H to identify his illocutionary intent pursuant to the
communicative presumption. The case of (34) brings this out clearly.
(34) If you're so smart, who is the voice of Bugs Bunny?
In order to identify the connection between the two clauses of this
sentence, it is necessary to supply several suppressed steps. A per-
Issues 228
formative paraphrase won't do the trick. Instead, the sequence lOoks
something as follows:
(34) a. If you're so smart (as smart as you say/think/suggest you are),
you know and can therefore tell me who is the voice of Bugs
Bunny.
b. To see if you can tell me ... , I'll ask you.
Conversational paraphrases can be provided analogously for the re-
maining cases of (35)-(37).
Once we recognize that not every conversationally usable sentence
must have a literal linguistic representation, or at least one that makes
literal sense, and that the requirement of conversational paraphrasabil-
ity suffices instead, we can forgo the wild-goose chase for deeper
structures from which to derive, by deletion, otherwise anomalous
sentences. Our approach is to regard such sentences as not used liter-
ally, in some cases as not even having literal meanings. However, they
are not sheer nonsense but readily decipherable ways of saying suc-
cinctly what could be said literally only in a cumbersome way. They are
readily decipherable because they are standardized in their illocution-
ary force. We have argued that this is not a matter of meaning, and we
have openly admitted that to take this position is to assume a fairly
clear-cut distinction between what is ungrammatical and what is other-
wise unacceptable. These distinctions impose clear limits on both the
power and the scope of linguistic explanation. But then linguistics can-
not do everything. We shall now see that leaving room for extra-
grammatical phenomena relieves linguistic theory of having to explain
the conversational role of other syntactic liberties and lexical omissions
besides those involved in illocutionary adverbials.
10.4. SYNTACTIC LIBERTIES
In linguistics as elsewhere there are few if any pure data. The data that
a theory has to explain, or at least accommodate, must be described
somehow, and how the data are described reflects at least low-level
theoretical commitments. Moreover, when a theory is constructed with
an eye to the data, the interests of simplicity and generality inevitably
require throwing out some of the data, not by ignoring them but by
explaining them away, either in terms of the theory in question or in
terms of some already accepted theory of phenomena that the recalci-
trant data fall under. When it comes to the point where residually
Standardization 229
stubborn data are excluded just to sustain the theory in question, it is
time to look for another theory.
Among other things, the object of a grammar of a language is to
describe a fluent speaker's intuitions of grammaticality. Ideally, every
string of words in a language is clearly and consistently judged by fluent
speakers to be either grammatical or ungrammatical, and those judged
grammatical constitute the set of sentences generated by an adequate
grammar of the language. In practice the grammatical/ungrammatical
distinction as reflected by speakers' intuitions is not clear-cut. Speakers
may be unsure in some cases; even when sure, they may disagree
among themselves; they may find some sentences relatively more
grammatical than others; and in different cases they may give different
reasons for judging sentences ungrammatical. A working grammarian
must take all this into account. If he retains the objective of producing a
grammar that generates just those sentences judged grammatical, he is
forced to throw out certain data and idealize those that remain. Or he
may opt for degrees of grammaticality and perhaps for kinds of relative
ungrammaticality. In this way he minimizes the unsalvageable data, but
if his theory is adequate, it will be at the expense of simplicity. Which-
ever tack he takes he cannot avoid (either in fact or in effect) putting
words in the mouths of his respondents, since he has to indicate to them
the kind of judgment he wants them to make, namely, of grammati-
cality or ungrammaticality.
What does all this have to do with a theory of speech acts? Sentences
and illocutionary acts do not correlate at all neatly. The semantics of a
language cannot coherently be based on correlating grammatical sen-
tences with illocutionary act types, since allowance must be made for
nonliteral and indirect illocutionary acts. Here we wish to point out
something further: an utterance does not have to be grammatical to
have been produced with an identifiable illocutionary intent. Instead of
uttering a grammatical sentence a speaker can produce a word, a
phrase, or a dependent clause and thereby successfully perform an
illocutionary act. For that matter, he can utter an ungrammatical sen-
tence that expresses his illocutionary intent. We maintain that any
theory is misguided which attempts to treat as grammatical (and to
assign semantic representations to) every locution that can be uttered
with an identifiable illocutionary intent. Even for those locutions that
native speakers judge to be in some sense acceptable, a theory of the
language must not automatically construe them as grammatical. 20 Even
if significant generalizations can be made about such diverse classes of
Issues 230
locutions as occur in sportscasts, commercials, pop lyrics, and on
traffic signs, medicine bottles, and cereal boxes, we think it a serious
theoretical mistake to regard as necessarily grammatical those locu-
tions that fall under such generalizations (see Sadock 1974, 139ft'),
Equally mistaken is requiring a theory of a language to single out all
those locutions, grammatical or not, that can be used to perform illocu-
tionary acts. Not only is a theory of grammaticality not a theory of
usability (if that is relevant to acceptability) but we believe there is no
hope at present for a theory of usability. Thus we maintain that linguis-
tic competence in the sense of knowledge of a language (assuming that
this is what the theory of a language captures) does not exhaust com-
petence at using expressions in a language. Chomsky remarks,
Use of language involves cognitive systems beyond grammatical and
pragmatic competence. The theory of performance, then, will attempt
to develop models incorporating grammar and other cognitive struc-
tures, as well as an account of the physical and social conditions of
language use that are ignored in the abstraction to grammar. (1977, 3)
Consider some examples of ungrammatical locutions whose utter-
ance is a readily identifiable illocutionary act.
(47) a. Close cover before striking.
b. No smoking
c. The Steelers going for a field goal
d. Lucerne two-ten low-fat milk
e. Slippery when wet
f. Two nonstudents, please
None of these locutions is a grammatical English sentence. Example
(47a) comes close to the sentence "Close the cover before striking the
match," and (47c) comes closer to being a sentence, needing only an
"are" before "going"; (47b) can be interpreted as "Smoking is not
permitted here," (47d) as "This carton contains Lucerne two-ten low-
fat milk," (47e) as "This road is slippery when wet," and (47f) as "I
would like tickets for two nonstudents." Although the examples are to
different degrees ungrammatical, at least by traditional standards, it is
no trick to decipher the illocutionary point of their use. The communi-
cative presumption is not waived merely by the utterance of an un-
grammatical locution. If there is a likely candidate for the illocutionary
intent of such an utterance, one may infer, barring any reason to the
contrary, that that candidate is the intent. To be sure, there are prec-
Standardization 231
edents for such utterances-some are even standardized-but that
does not make them grammatical. Rather, it makes their illocutionary
intents especially easy to identify.
These examples illustrate how speakers can use ungrammatical sen-
tences or even nonsentences with identifiable illocutionary intents. To
use a locution thus is to take what we call a syntactic liberty. A great
deal of ordinary language use involves syntactic liberties, everything
from answers to questions to newspaper headlines and telegrams. We
do not intend to cover the subject in any great detail, but we should
give some indication of how this phenomenon fits into the speech act
schema. The problem is that if the locution uttered is not a grammatical
sentence or not a sentence at all, it has either no meaning or not enough
meaning to make an utterance of it a locutionary act of the form, saying
that *( .. .p .. . ); so the SAS is blocked at the level of saying. But there is a
solution. Consider slips of the tongue (or mispronunciation). If S acci-
dentally utters something other than what he intends to utter, H may be
able to figure out what S meant to say. If S utters "Did you hear about
the First Lady's girl bladder operation?" presumably he meant to ask
about her gall bladder operation, and H makes the appropriate infer-
ence. H makes a similar inference in deciphering malapropisms, where
S uses a word to mean something that some similar-sounding word
means, like "resemble" for "resent" in "I resemble that remark." The
same thing occurs in the case of syntactic liberties, except there is
nothing unintentional about the locution used. H, assuming the CP to
be in effect, is able to figure out what S means by what he utters,
perhaps by associating a grammatical sentence with the locution ut-
tered. 21 Perhaps what one does, using strategies developed through
experience, is ascertain *( ...p ... ) directly, without the mediation of a
grammatical sentence. In any event, our point is that the SAS can be
easily stretched to accommodate whatever goes on-we offer no de-
tailed account-when illocutionary acts are successfully performed
with the use of ungrammatical sentences or nonsentences.
Another kind of syntactic liberty is worth mentioning. 22 In these
cases a grammatical sentence is used, but what the speaker means by it
is not what the sentence means, at least not exactly. Examples (48) and
(49) illustrate, respectively, what we will call scope shifting and lexical
omission.
(48) Walter only watches the news.
(49) The Steelers don't look tired, they are tired.
Issues 232 Standardization 233

Normally (48) is used to say that Walter watches only the news, but f. I thought the side window was down but it was up . . . as I
literally S says that Walter does nothing else but watch the news (or found when I put my head through it.
if "watches" is stressed, that Walter does nothing else with the news g. My car had to tum sharper than necessary owing to an invisible
but watch it). Words like "only," "just," "also," and "even" exhibit lorry.
scope shifting. In (49) it is literally denied that the Steelers look tired,
but what a speaker of (49) means to say is that the Steelers don't merely
look tired, they are tired. Other examples of lexical omission include
"or" for "or else" and "and" for "and then." To think sentences like
(48) and (49) mean what their users intend them to mean as well as what
they literally mean but are not intended to mean is to multiply meanings
beyond necessity.
Our general policy is not to take as a matter of meaning whatever can
be construed as a matter of communicative intention or inference. Thus
a locution that violates grammatical rules need not be regarded as
grammatical just because it can be used with identifiable illocutionary
intent. Similarly, we need not attribute additional meanings to sen-
tences that are standardly used nonliterally, at least in ways like those
involving scope shifting or lexical omission (obviously, we do not want
to rule out sentential idioms like "That's water under the bridge"). We
believe that the interests of theoretical simplicity are better served by
dividing the labor between a narrowly conceived grammar of a lan-
guage and a pragmatic account of the use of sentences and other locu-
tions. 23 Although generalizations can undoubtedly be made about the
various types of syntactic liberty (restrictions on scope shifting, for
instance), we see no reason to regard them as facts about linguistic
competence rather than as facts about standardized linguistic perform-
ance. We have said little about how these phenomena take place, that
is, about how speakers and hearers produce and perceive locutions
other than grammatical sentences (and grammatical sentences used
other than in their strictly grammatical way) with identifiable illocu-
tionary intents. Whatever the explanation ultimately looks like, it must
account for the fact that people do say and do understand remarks like
the following descriptions of automobile accidents by drivers: 24
(50) a. The other man altered his mind, so I had to run over him.
b. A pedestrian hit me and went under the car.
c. I badly misjudged a woman crossing the street.
d. I blew my hooter (hom) but it would not work because it had
been stolen.
e. Coming home I drove into the wrong house and collided with a
tree that I haven't got.
 Speech Act Schema and Psychology 235

Chapter Eleven The Speech Act Schema I I

and Psychology SPEAKERS
I I
Speaker S's pragmatic intent (PI)
I

I Speech Production Mechanism I

Sentence e is chosen

I Articulatory Apparatus I
In our introduction we mentioned a venerable conception of linguistic 1
communication that dates back at least to John Locke. Linguistic t
Acoustic signal (sentence) e
communication is viewed as a process whereby a speaker encodes a
message into a signal from which a hearer decodes the message. As far
as it goes this picture is probably correct, but it idealizes linguistic
communication in various important respects. 1 The utterance is taken
to be unambiguous and no mechanisms for contextual (linguistic or
Speech Recognizer
nonlinguistic) disambiguation are required. The speaker is presumed to
be speaking literally, not ironically or sarcastically, and so is taken to
mean just what he says. The reference and force of the utterance are Comprehension 1: Sequence of phones
taken as uniquely determined by the meaning (or message). The
speaker is assumed to be speaking directly and not indirectly. As part I Lexicon and Surface Processor
showed, theoretical devices must be added to this idealized conception
if we are to account for the full variety of linguistic communication Comprehension 2: Surface structure
represented in the SAS; when this is done the process will have the
general structure represented in figure 11.1.
However, if construed as representing the actual process of linguistic Syntax, Semantics, Contextual Beliefs
communication, the model is seriously defective in failing to reflect the
parallel or simultaneous nature of much of the information processing. Comprehension 3: Operative sentence meaning: understanding e
Clearly, different levels of processing are going on simultaneously as
we speak or hear a sentence from "left to right" (see Marslen-Wilson Pragmatics, Contextual Beliefs
1975). At the very least, while a hearer is determining the meaning and
I: reference of some earlier part of an incoming sentence, he is also de-
Comprehension 4: Understanding S: recognizing CI
termining the syntax of some later part and the phonetics of some still
later part. However, the model does depict the overall direction of
information flow, and even though constituents of e may proceed Figure 11.1 Some stages of information flow in linguistic communication
through the steps at different times, we will use this diagram as a
starting point for our investigation.
Issues 236
Semantic and pragmatic information enter at two prominent points:
(a) at the beginning of the speech production phase, and (b) at the end
of the speech comprehension phase. Although much more work has
been done on comprehension than on production, some interesting
preliminary findings about production are worth considering.
11.1. PRODUCTION
The most interesting proposal to date for a model of speech production
comes, we feel, from the work of Fodor, Bever, and Garrett (1974) and
Fodor (1975). Their proposal has three central features. First, speech
production is taken to be a special case of considered action and as
such falls under any plausible psychological model for considered ac-
tion. It is proposed, second, that the best general model for considered
action is decision-theoretic and, third, that the performance of such
speech acts involves translating pragmatic intentions, formulated in a
language of thought, into sound sequences.
Issues surrounding the existence and nature of a language of thought
are highly controversial;2 we will assume only that pragmatic intentions
are formulated in some system of representation. 3 Given this assump-
tion, we will consider the ideas that speech production is a species of
considered action and that speech production (speech acts) involves a
translation of a pragmatic intention into a signal.
Fodor, Bever, and Garrett (1974, 375) offer the following schematic
model of relevant information processing underlying ideal4 cases of
speech production:
(1) A variety of candidate messages are formulated in mentalese ... (2)
The speaker attempts to predict the consequences of communicating
one or another of the candidate messages . . . (3) The utility of bringing
about these various [effects] is [calculated] and compared ... (4) A
best candidate is chosen. (5) [This message is translated into the] sen-
tence which best expresses [it]. (6) The sentence is uttered.
This model is an idealization in two ways, according to Fodor, Bever,
and Garrett. First, steps (2) -(4) are not always present in that we do
not always consider the consequences of what we communicate nor do
we consider a variety of possible messages. Second, we do not always
choose the best way of saying something, so step (5) is not always
present. However, we sometimes do consider how to phrase what we
want to communicate, and there is no obvious reason why we should
not take this proc.ess to be of the same kind as the process marked by
Speech Act Schema and Psychology 237
steps (2)-(4). Thus step (5), when present, could have internal struc-
ture like the following:
(a) The message is translated into a variety of sentences.
(b) The consequences of uttering each of them are calculated and com-
pared.
(c) The best sentence for achieving S's pragmatic intent is chosen.
Two serious idealizations still remain unnoted. One is that our prag-
matic intention often comprises more/less than communicating some
message to H. Sometimes we have perlocutionary intentions as well
(see section 4.5), and sometimes we are intending to perform some
conventional speech act (see chapter 6). In either case S's pragmatic
intention will not be Gust) to communicate some message in uttering
the chosen sentence. When S does intend Gust) to communicate some
message, the relation between that communicative intent and the
chosen sentence is rarely one of translation, even though from S's point
of view disambiguation is not a problem: there is reference to be fixed
and illocutionary force to be delimited (see chapter 2). The closest
approximation to a translation of a given communicative intent would
be an "eternal" sentence (see Quine 1960, sec. 40) containing an ex-
plicit performative prefix, and we rarely speak in this fashion. The
relation between the communicative intent (the message) and the
meaning of the chosen sentence is closer to compatibility (discussed in
chapter 2) than to translation. For nonliteral illocutionary acts, clearly
the relation between the communicative intent and the meaning of the
sentence uttered will be less direct than with literal utterances. In short,
if we are correct in part I, this schematic model of a speech producer
will more and more approximate the outlines of the elaborated SAS
presented in chapter 4, and so the speaker's pragmatic intention will,
depending on the case, comprise various aspects of the SAS.
A second remaining idealization concerns the determination of the
pragmatic intention (PI). 5 Since a speaker's PI can include a number of
subintentions (in particular, intentions to perform utterance, locution-
ary, illocutionary, and perlocutionary acts), part of what goes into
determining a speaker's PI will be such familiar things as mutual con-
textual beliefs, beliefs about H's beliefs, beliefs (and desires) about the
social and physical context, and S's beliefs and desires concerning the
nature and direction of the discourse. If these and the previous remarks
are correct, our schematic speech production model should have the
following minimal structure: 6
Issues
Speech Production Model (SP):
238
1. Speaker S has a variety of beliefs and desires concerning such fac-
tors as:
(a) the nature and direction of the discourse,
(b) the social and physical context of the utterance,
(c) H's beliefs in general, H's beliefs pertinent to S's impending
remark in particular, and whatever mutual contextual beliefs H
shares with S.
2. On the basis of 1, S forms a variety of pragmatic intents
PIt, ... , PIn, which may include subintentions:
(a) to utter something (utterance intent),
(b) to say something (locutionary intent), and so
(c) to refer to something (referential intent),
(d) to perform some illocutionary act(s) of a communicative or con-
ventional sort (communicative intent, conventional intent),
(e) to perform these acts literally, nonliterally, directly, or indi-
rectly,
(f) to have various effects on the hearer H (perlocutionary intent).
3. S attempts to predict the consequences of fulfilling these various
intents.
4. The utility of each is assessed.
5. A particular pragmatic intent PIi is formed.
6. A variety of expressions et, ... , en are constructed, each compati-
ble with PIi.
7. S attempts to predict the consequences, for fulfilling S's pragmatic
intent PIi, of uttering each expression.
8. The expression ei judged most likely to succeed in fulfilling PIi is
chosen.
9. Expression ei is uttered.
We are not aware of any experimental work directly relevant to steps
1-5. However, there do seem to be data relevant to parts of steps 6-9.
First, the work of Tannenbaum and Williams (1968) and Osgood
(1971) suggest that there are regular connections between the form of
the chosen expression ei and aspects of the speaker's pragmatic intent;
thus it is pertinent to steps 7 and 8. Tannenbaum and Williams gave
subjects a paragraph to read concerning some topic such as (1) trains,
(2) cars, or (3) something else. The subjects were then given a picture
of, for instance, a train hitting a car, with an A (for active) or a P (for
passive) written in one comer. Subjects were then asked to produce the
active sentence "The train is bitting the car" or the passive sentence
1
I
Speech Act Schema and Psychology 239
"The car is being hit by the train," depending on which letter occurred
in the picture. The resulting latencies to completion had the structure:
A: (1) < (3) < (2)
I P: (2) < (3) < (1)
:
!
If the topic of the paragraph was referred to by a certain NP, subjects
were best able to continue mentioning th&t topic with sentences having
that NP in surface subject position, even if a passive sentence (which is
longer) was required to get it there. Osgood (1971) illustrated a similar
tendency among subjects asked to describe a short skit in simple sen-
tences. As an object became the focus of attention, NPs referring to
that object tended to occur in surface subject position. 7
Second, there is a body of data relevant to aspects of the process
underlying step 6. If the view of linguistic communication embodied in
the SAS is right, knowledge of the language plays a major role in
enabling H to recognize S's communicative intention. Fodor (1975,
106) proposes that this knowledge be explicated in terms of shared
linguistic conventions:
One might think of the conventions of the language as a sort of cook-
book which tells us, for any C that can be communicated by an expres-
sion of the language, "if you want to communicate C, produce an
utterance (or inscription) which satisfies the descriptions D 1 , D 2 ,
Dn" where specimens Ds might be syntactic, morphological, and
phonological representations of the utterance. The converse remarks
hold for the hearer: To know the conventions of a language is at least to
know that an utterance which satisfies D 1 , D 2 , Dn also standardly
satisfies the description "produced with the intention to communicate
C."
This suggests (Fodor 1975, 109) that it is with these descriptions
D 1 , Dn that contact is made with a grammar of the language:
A generative grammar of L specifies (some or all of) the descriptions
that a token must satisfy if it is to conform to the linguistic conventions
of L. To put the same point slightly differently, it specifies, for each M,
the descriptions (morphological, phonological, syntactic, etc.) that a
token must satisfy if it is to belong to that sentence type which ex-
presses M in L.
Given that messages (contents of a communicative intention) are rarely
identical to the meaning of the sentence uttered, is there evidence that
any of the linguistic levels between message and signal (semantic rep-
resentation, deep structure, surface structure) are or are not computed
Issues 240
S are mediated by knowledge of language use in conjunction with beliefs
~VP
NP
A
Det N
I pects of the processor can assign enough of a syntactic description to
the
Figure 11.2 A partially developed clause, constructed top to bottom and left to
right
during the process of speech production, and if they are, what might
the nature of the operations underlying their computation be? Accord-
ing to Fodor, Bever, and Garrett (1974, 434) there are considerable data
in favor of the view that surface structures are one such level and that
the process of constructing a surface phrase marker is from top to
bottom, left to right, and clause by clause, thus yielding structures in
memory having the form of the tree in figure 11.2. 8 Though these results
are psychologically interesting in their own right, they shed little light
on the semantic and pragJIlatic aspects of speech production. As Fodor,
Bever, and Garrett (p. 397) remark, "Both the conceptual and empiri-
cal issues in the most interesting areas of the production problem are
largely unsolved." The picture does not change much in a recent sur-
vey by Clark and Clark (1977, 248): "It is fairly clear what consider-
ations speakers must pay attention to, but it is far from clear what
mental processes are involved or how the final decision is arrived at.
The study of sentence planning has barely begun." Since it is the
semantic and pragmatic aspects of speech we are primarily interested
in, we tum to the body of literature surrounding speech perception and
comprehension, where there seem to be fewer methodological prob-
lems and firmer results. 9
11.2. COMPREHENSION
Speech comprehension can be divided into four steps, as represented in
figure 11.1. Phonological ability and lexical knowledge take the hearer
from step 1 to step 2, and semantics (along with syntax and contextual
beliefs) mediates steps 2 and 3. We will focus on steps 3 and 4, which
Speech Act Schema and Psychology 241
about context. In order to understand this stage, it is useful to break
steps 3 and 4 down into several substeps.
To reach step 3, H must settle on the operative reading of the ex-
pression; to do this he must first assign the right syntactic descrip-
tion, then assign the right meanings to the lexical items, and finally
assign the right meaning to the sentence. Given that the syntactic as-
get the rest going, the process of speech comprehension then resolves
into two major subproblems. We call this the operative meaning stage
of comprehension:
3. Operative meaning stage:
(a) Determine the operative lexical reading(s) of expression e.
(b) Determine the operative sentential meaning(s) of expression e.
In order to reach step 4, H must infer the speaker's communicative
intention, CI. This process resolves into two stages, the propositional
content stage and the illocutionary force stage, each comprising major
subproblems:
4. Propositional content stage:
(a) Determine the intended referents.
(b) Determine the propositional content.
Illocutionary force stage:
(c) Determine what direct (literal or nonliteral) illocutionary inten-
tion S had in uttering e.
(d) Determine what indirect illocutionary intention S had in utter-
ing e.
(e) Determine S's communicative intention, CI, on the basis of 4(c)
and 4(d).
A hearer who manages to get from 3(a) to 4( e) has in effect completed
his side of the process of linguistic communication. To better under-
stand this process we examine the psychological details of these stages.
Since we process a sentence in real time from "left to right" (speak-
ing metaphorically, as if of written English), and since we start pro-
cessing almost immediately, it is probable that we process almost all of
this information in parallel. That is, syntactic analysis may be influ-
enced by semantic analysis and semantic analysis by pragmatic anal-
ysis-which may in turn be influenced by prior syntactic analysis. So
Issues 242
we should not view these steps as strictly ordered, either for the sen-
tence as a whole or for its constituents. Rather, the ordering reflects
general information flow and the necessity of discussing components
one at a time.
11.2.1. Operative Meaning Stage
Step 3(a) involves determining the intended readings of the minimal
functioning constituents of the expression. Step 3(b) involves integrat-
ing these into a reading for the whole sentence and determining its
intended reading. Clearly this is, in real time, a highly interactive par-
allel process, from which these steps are an abstraction. Another di-
mension of abstraction is away from syntax. Semantic processing can
serve to assist in syntactic processing and vice versa. Nevertheless, in
discussing the lexical level we will try to discuss problems in the
semantic representation of lexical items independently of syntactic
details other than clause boundaries.
Let us assume that words are the minimal semantic units involved in
speech comprehension. Then the hearer goes through two substages:
he must represent the potential meanings of each word and determine
its operative meaning.
Lexical Representation during Comprehension
We said in section 8.1 that there were some considerations in favor of
replacing lexical definitions with inference rules in an empirically ade-
quate semantics. We left aside the question of whether such definitional
processes were part of the comprehension process. We now tum to
some of the relevant psychological data.
Fodor, Fodor, and Garrett (1975) present both intuitive and experi-
mental evidence against the hypothesis that lexical decomposition is
usually involved in comprehension. On the intuitive side, the hypothe-
sis predicts that the more elaborate the definition, other things being
equal, the more complex and difficult the comprehension process
should be. In particular, when the definition of one word is a proper
part of the definition of another word, there should be asymmetry in the
difficulty of comprehending them. Yet pairs like the following do not
seem to exhibit noticeable asymmetries: 10
(1) a. x is unmarried: x is a bachelor
b. x chases y: x catches y
Speech Act Schema and Psychology 243
On the experimental side, Fodor, Fodor, and Garrett report an ex-
periment in which subjects were required to evaluate the validity of
different arguments such as:
(2) a. If practically all of the men in the room are not married, then few
of the men in the room have wives.
b. If practically all of the men in the room are bachelors, then few of
the men in the room have wives.
Each of the arguments contained either a quantifier (" all," "few") or an
explicitly negative (free) morpheme ("not") together with one of the
following sorts of negative elements: (EN) another explicitly negative
free morpheme; (MN) a morphological negative, "in-," "un-," "im-";
(IN) an implicitly negative morpheme, "doubt," "deny"; or (PDN)
a pure definitional negative, "bachelor," "kill." The importance of
PDNs is that they contain negation at the semantic level but have no
syntactic reflexes of negation. If semantic representations are decom-
positional and computed during comprehension, there should be no
significant difference between the reaction times to a correct evaluation
of arguments containing PDNs, such as (2b), and times for the evalua-
tion of the others, such as (2a). Since it is fairly well established that
when negatives interact with quantifiers or other negatives, latencies
are lengthened, this experiment can be seen as testing for the presence
of a negative element in the comprehension of a word like "bachelor."
According to Fodor, Fodor, and Garrett, arguments containing
PDN s (2b) were significantly easier than paired arguments with ENs
(2a). In addition, the difference in reaction times between ENs and
PDNs was significantly greater than the difference in times between
either ENs and MNs or ENs and INs (though figures were not given).,
That is, EN - PDN > EN - MN, and EN - PDN > EN - IN. Fodor,
Fodor, and Garrett conclude: "We take this result to suggest strongly
that PDNs do not act as though they contain a negative element in their
linguistic representation; and therefore, that PDNs are not semantically
analyzed at any level of linguistic representation" (p. 522).
This experiment is hard to evaluate in the absence of more details
(see Katz 1977b), but two considerations would have to be ruled out
before we could accept it as a strong case. First, it was also found that
subjects performed the same for MNs and INs, and this might argue for
decomposition of these items. 11 Second, a decompositionalist might
claim that the results show only that decomposition during comprehen-
 Issues 244
Sentence Semantic ~ Inference Entailments
Signal ~ Understander ~ Representation Rules ~ of Signal
Figure 11.3 Two-stage model of sentence comprehension
sion did not reach negation. For example, the relevant decomposition
of "bachelor" might simply be (Def) bachelor: ADULT & MALE &
SINGLE. If such a theory were supplemented by the inference rule
SINGLE -? NOT-MARRIED, which was optional in the process of com-
prehension, then the experiment would prove nothing. 12
Another consideration Fodor, Fodor, and Garrett cite in favor of
inference rules over lexical decomposition is comprehension time (p.
526). If words like "bachelor" and "kill" are not decomposed during
the process of understanding, then "bachelor" is the minimal meaning-
contributing (or meaning-representing) unit-or its translation in the
comprehension language is (see Fodor 1975, 150-152). In that case the
vocabulary of the representation would approximate that of the lan-
guage itself. This means that comprehension could, semantically speak-
ing, be direct and quick since the process of drawing inferences could
be separated out as a distinct stage (see figure 11.3). As Fodor (1975,
151) comments, "the operations of the sentence understander are on-
line operations. We understand an utterance when we hear it. But the
operations of the logic may take any amount of time at all." Is quick-
ness a virtue? Marslen-Wilson (1973) reports the results of an experi-
ment involving the restoration of disrupted words during a sentence
shadowing task. In the course of arguing for a parallel processor with
i
I,
I
high interaction between phonological, morphological, syntactic, and
semantic information processing, Marslen-Wilson notes that good
shadowers working at 250 msec can be seen to be affected by semantic
information; indeed, at a shadowing latency of 250 msec, they began to
repeat the target-words when only the first syllable could have been
heard.
The model diagrammed in figure 11.3 suggests that sentences receive
a quick semantic representation, that quickness here is a virtue. But
then we must ask what exactly are these semantic representations (or
formulas in mentalese) if understanding "bachelor" need not involve
inferring that the referent is male or adult or unmarried? What does
understanding "bachelor" amount to if it has nothing to do with these
other notions? Perhaps one could say that the semantic representation
Speech Act Schema and Psychology 245
does no more than represent the property of being a bachelor and that
what constitutes that property is inferred. Then, of course, we need an
account of properties as well as an account of representation which
explains understanding, and we are a long way from this (but see Field
1978).
Ambiguity
On hearing an expression, sometimes we are aware that it is ambig-
uous- we may even have each meaning in mind. At other times we are
not aware of the ambiguity; either we do not know one of the meanings
or one of them fails to come to mind. An adequate theory of com-
prehension must account for both kinds of cases.
A variety of psychological work has been done on three sources of
ambiguity:
Lexical: I found a bat.
Surface structure: It was in a little bat house.
Underlying structure: He had the bat stolen.
There is some evidence (see MacKay and Bever 1967; MacKay 1966)
that when ambiguity is perceived, lexical ambiguity is perceived quick-
est, then surface ambiguity, and finally underlying ambiguity. In Mac-
Kay (1966) subjects wer~ visually given sentence fragments containing
these types of ambiguity, separately and jointly, and asked to complete
the sentences by saying the entire sentence out loud. The time for
completion was recorded and the subjects were asked whether they
had noticed the ambiguity. If they had, their responses were put aside.
The results are as indicated in figures 11.4 and 11.5. Clearly something
different is going on in each of these cases. MacKay (1973) has sug-
gested that lexical and surface ambiguities are processed (by some
finite state device) in short-term memory, but that underlying am-
biguities must await transfer to long-term memory before they can be
processed. However, this division (apparently) conflicts with the pres-
ently most plausible account of disambiguation, the clausal-closure
theory discussed in the next section. On that theory, clauses are pro-
cessed in short-term memory, then recoded for long-term memory. But
this computation in short-term memory requires deep structure pro-
jections, and so does processing transformational information. If both
theories were correct, there would, contrary to fact, be no such thing as
underlying ambiguity, since by MacKay's account such computation
must await long-term transfer whereas by the clausal-closure theory
Issues 246 Speech Act Schema and Psychology 247

L - Lexical Ambiguity
S - Surface Ambiguity
12.0 U - Underlying Ambiguity

10.0 (a) Ambiguities at the same (b) Ambiguities at

Ambiguous sentences linguistic level different linguistic levels
o Corresponding '00'
non ambiguous u 11.0
sentences ~
tl)

~
.
'"
u !=I
~ .9
tl)
9.0 ~p.. 10.0
. S
0
!=I u
.g
tl)
!=I
ro
"a :.a
tl)
S0 ~
u 9.0
!=I
ro
:.a
tl)

~
8.0
8.0~__~~____~~__~~~__~__~~____~~____~_____
L-L S-S U-U L-S L-U S-U

I
0

t t t t
Type of Ambiguity

Figure 11.5 The median completion time for the six types of multiple ambigu-
ous sentence fragments, with ambiguities at (a) the same linguistic level and at
(b) different linguistic levels. (From MacKay 1966)
7.0
Lexical Surface Underlying Multiple
what goes to long-term memory has been unambiguously recoded. It is
Type of Ambiguity
not clear at present how to reconcile the findings supporting each of
these theories, but the evidence supporting MacKay's account is
Figure 11.4 The median completion time for multiple ambiguous sentenc~ frag-
sometimes weak.13
ments for the three types of ambiguous sentence fragments and for therr cor-
responding nonambiguous fragments. (From MacKay 1966)
Disambiguation
What goes on when we hear an ambiguous expression which, though
we know on reflection to be ambiguous, we do not at the time perceive
to be ambiguous? And, what happens when we perceive an ambiguity
but immediately pick one meaning as the operative intended one? Is the
former process like the latter, only unconscious?
One of the more promising hypotheses concerning the general con-
straints on disambiguation (Garrett 1970; Lackner and Garrett 1973;
Bever, Garrett, and Hurtig 1973) postulates three information process-
ing stages:
Issues 248
Clausal-closure Hypothesis:
Stage 1: H computes clause-internal ambiguities.
Stage 2: At the end of a clause, H picks one meaning for the whole
clause and continues.
Stage 3: If that meaning turns out to be inappropriate, H goes back to
stages 1 and 2.
There is evidence for each of these stages.
Stage 1. In a dichotic listening task Lackner and Garrett (1973) pre-
sented subjects with an ambiguous sentence such as (3) in one ear.
(3) The spy put out the torch as our signal to attack.
To the subject's other ear they presented (at a 5- to 10-db lower inten-
sity level) either a neutral context sentence or a disambiguating (or
biasing) context sentence such as (4).
(4) The spy extinguished the torch in the window.
Subjects were instructed to attend to what they heard in the ear pre-
sented with the ambiguous sentence and to begin paraphrasing it before
the sentence ended. Even though after the experiment subjects could
not give any information about the material in the unattended ear,
Lackner and Garrett found (pp. 365 - 366) that the unattended biasing
sentences significantly affected the choice of paraphrase. They argue
from their data (p. 370) that some linguistic analysis of the unattended
sentence is taking place, since occasionally the biasing sentence, to
exert its effect, had to be analyzed beyond the lexical level to the
phrase level, as in the following examples:
(5) The sailors liked the port at night. (ambiguous)
(6) The sailors liked to be in port at night. (biasing)
(7) Visiting relatives can be a bore. (ambiguous)
(8) I hate relatives who visit often. (biasing)
They conclude that while a subject is listening to an ambiguous sen-
tence and determining its meaning "both readings are in some sense
available to him. If this were not the case, it would have been impossi-
ble to bias the interpretation of the ambiguous sentence" (p~ 367).
Although this may be true, the occurrence of any of several different
processes is compatible with their claim. As they note (p. 361), the
subject can compute all the options for the sentence when an ambiguity
is encountered or he can postpone assignment until context can be used
Speech Act Schema and Psychology 249
to determine the appropriate reading. Lackner and Garrett (p. 371)
prefer the multiple ambiguity hypothesis because it can explain the
effect of context on disambiguation and because a parallel processing
strategy can help explain why there should be differences in processing
difficulty between ambiguous and unambiguous sentences. Probably
most would agree that these are not overwhelming considerations in
favor of parallel processing of ambiguities, but in the absence of a clear
alternative, it should be given the nod (see also Holmes, Arwas, and
Garrett 1977).
Stage 2. Evidence in favor of stage 2 comes from Bever, Garrett, and
Hurtig (1973). On their view the clause is a primary perceptual unit
whose elements perceptual operations map directly onto underlying
structures. When a clause boundary is reached, its contents are re-
coded in a fairly abstract form, outside of immediate memory, leaving
immediate storage free for the next clause. Ambiguities are computed
during the projection onto underlying structures, but at the point of
clausal recoding, one projection is selected and the rest are dropped.
Bever, Garrett, and Hurtig have various kinds of support for this the-
ory. One kind of support arises from a reanalysis of the data in MacKay
(1966). In this reanalysis they grouped MacKay's sentences into two
classes: those, like (9), that could be complete clauses at the underlying
levels and those, like (10), that could not be:
(9) Although I knew the new position had advantages . . .
(10) After her injury that summer she couldn't bear ...
MacKay's data showed no effects of ambiguity for sentences of the first
type, but effects were found in sentences of the second type. On Bever,
Garrett, and Hurtig's theory these results follow from the fact that just
before a clause boundary is the time of maximum processing load and
hence of the potential effects of such a load, whereas after a clause
boundary one reading has been selected and there is little processing
effect.
To test this idea further Bever, Garrett, and Hurtig (1973, experiment
2) had subjects complete ambiguous incomplete sentences such as the
following. Notice that some of the incomplete sentences had complete
internal clauses and others did not.
(11) a. After taking the right turn at the
b. After taking the right turn at the intersection
c. After taking the right turn at the intersection I
Issues 250 Speech Act Schema and Psychology 251

d. After taking the left turn at the Table 11.1

e. After taking the left turn at the intersection
f. After taking the left turn at the intersection I
Clearly the effects of ambiguity on the second class (those with in-
complete internal clauses) should have been considerably greater than
on the first. The results are curious in that only effects of underlying
ambiguity were significant, so the theory is supported only to that
extent. Bever, Garrett, and Hurtig propose that the classification of
ambiguities usually given is not based on the perceptually most salient
characteristics. Perhaps ambiguities should be classified in terms of the
perceptual independence of the operations recovering underlying rela-
tions. They suggest that "it is the perceptual independence of interpre-
tations that governs behavioral differences in response to ambiguities"
(p. 285). If the operations required at a given level are quite different
from one another, computational difficulty goes up before clause
boundaries.
Support for this idea, as well as further support for stages 1 and 2,
comes from another experiment (experiment 1) reported by Bever,
Garrett, and Hurtig (1973), in which they presented subjects with sen-
tences (both ambiguous and nonambiguous) such as the following:
(12) a. Be sure that you take the right turn. (lexical ambiguity)
b. Be sure that you take the left turn.
c. Be sure that you take the correct turn.
(13) a. The paper presented carefully limited analyses of the problem.
(surface ambiguity)
b. The paper presented today limited analyses of the problem.
c. The paper presented very limited analyses of the problem.
(14) a. The shooting of the Indians bothered the agent. (underlying
ambiguity)
b. The shooting from the Indians bothered the agent.
c. The shooting at the Indians bothered the agent.
The subjects were instructed to interpret each sentence and then add
another sentence to complete the story. The results are indicated in
table 11.1. The figures on the last line are particularly interesting in that
they show subjects responding faster to ambiguous sentences than to
nonambiguous sentences, and in the case of underlying ambiguity the
difference was significant. That there should be no significant increase
in completion time is predicted by the clause-boundary hypothesis,
Time (seconds) to start the responses to ambiguous and unambiguous
sentences in experiment 1
Type of Ambiguity
Smface Underlying
Lexical Structure Structure
Ambiguous sentences 5.28 9.03 8.83
Corresponding unambiguous 4.93 9.40 9.72
sentences
Mean difference of responses +0.35 -0.27 -0.89
(ambiguous - unambiguous)
Percentage of Ss responding faster 34% 56% 82%
to ambiguous versions
Source: Bever, Garrett, and Hurtig 1973.
since subjects were instructed to wait until they had interpreted the
sentence before continuing. But the finding that underlying ambiguities
facilitated completion was unexpected. On the theory that all ambi-
guities are simultaneously but independently computed up to clause
boundaries, since' either interpretation of each sentence was acceptable
in the experiment, the chances of computing an interpretation faster are
actually increased by having more available readings, not decreased. 14
This is on the assumption (p. 280) that once a reading is consciously
perceived, further processing on that sentence stops. How can these
effects be reconciled with the facts recorded earlier from MacKay
(1966) indicating that underlying ambiguity had a retarding effect on
completion times? Again the idea of classifying ambiguities by percep-
tual differences is useful. Suppose we give sentences (12) -(13) Bever,
Garrett, and Hurtig's fairly standard transformational analyses as indi-
cated in figures 11.6 to 11.11. One can see that the lexically ambiguous
sentences have identical underlying structures, and the superficially
ambiguous sentences have underlying structures with "a minor differ-
ence, characteristically in placement of a modifier" (p. 282), However,
for the sentences involving underlying ambiguity, the deep structures
differ in important underlying relations such as subject and object, and
the ambiguities cannot be resolved by surface bracketing. In the case of
lexical ambiguity (12a) the perceptual operations used to compute the
readings are virtually identical, in the sur(ace ambiguity case (13a) they
are very similar, but in the deep structure cases (14a) they are quite
distinct. Thus the theory predicts that the last case will exhibit more
Issues 252 Speech Act Schema and Psychology 253

S S

~ ~
Imperative
NP VP
Imperative
NP VP

VP S VP S

V
/\ Adj NP VP V
/\ Adj NP VP

I I /~ I I
be sure that
/~
you
V NP
be sure that you
V NP

I~ take
I~
Det N S
takeDet N S

II~ II~
the
turn NP VP
the
turn NP VP

/\~
Det N V Adj
/\A
Det N V Adj

I I isI I I I I I
the turn right} the turn is {right }
{ left correct

Figure 11.6 Structure underlying (l2b) and one reading of (12a) Figure 11.7 Structure underlying (l2c) and one reading of (l2a)

interference characteristics when tested before a clause boundary, and
that is what was found. (See Bever, Garrett, and Hurtig 1976, 219-220,
for further discussion.)
A question left unsettled (Bever, Garrett, and Hurtig 1973, 285) is
whether the operative notion of a clause should be that of a surface or
underlying clause. Intuitively, one would think that surface structure
clause is the relevant notion. One reason is that it may be necessary to
recover more than one underlying clause to interpret a surface clause,
and if short-term memory were emptied at the first deep clause, ambi-
guity at the underlying level could not be perceived. Since such ambi-
guity is perceived, the surface clause must be the relevant notion of a
clause, though an exact definition of this notion is far from clear. Is As
Carroll (1978, 506) remarks, "In sum, the identity of those sentence
perception units remains elusive."
Stage 3. Evidence for stage 3 of the clausal-closure hypothesis comes
from data in both the previous experiments and new ones, as well as
from introspection. We have all had the experience of taking one clause
in a sentence one way only to encounter a word that forced us to
reinterpret the sentence from the beginning. This suggests that we have
retained only one reading R1 and must compute the alternative reading
R2 the second time around. The problem is to devise a test situation
that will distinguish computing R2 then and there from retrieving R2
from temporary store. It is not clear to us that either the study by Foss,
Bever, and Silver (1968) or that by Carey, Mehler, and Bever (1970),
which are usually cited as supporting the single-reading theory, actu-
ally do distinguish these two possibilities. More recent work supporting
a single-reading theory also has problems. For instance, Schvaneveldt,
Meyer, and Becker (1976) used a wordlnonword sorting task to support
Issues 254 Speech Act Schema and Psychology 255

S S

~VP
NP
/~VP
NP

/\
Det N V Adv
/\
Det N V

I I I I I I I
the paper presented {CarefUllY} NP the paper presented NP
today

N S S
N S S

I ~VP ~VP I
analyses
~VP
NP
~VP
NP
analyses NP NP

I /\ I /\ I
N
~
Adv
V Adj N
I
V
A Complement
N V Adj N V Complement

I I I I I A I I I I I I A
analyses are {carefUllY} limited analyses are of the problem
analyses are limited analyses are of the problem very
Figure 11.8 Structure underlying (13b) and one reading of (13a) Figure 11.9 Structure underlying (13c) and one reading of (13a)

a single-reading theory, but they did not control for word and meaning
frequencies. Tyler and Marslen-Wilson (1977) argue for a single-reading
account as well but fail to control for syntactic cues to disambiguation.
We have not yet said anything about exactly how one reading for an
ambiguous sentence is, in fact, computed and selected. This problem
arises at the end of each clause, and at the end of sentence. At both
points factors relating to the plausibility of an interpretation, including
its coherence with antecedent and expected events, seem to playa role.
To determine experimentally how this works would involve testing for
how beliefs, and especially mutual contextual beliefs, affect the com-
putation and selection of alternative readings, and how these beliefs
interact with various syntactic strategies (see Fodor, Bever, and Gar-
rett 1974, 328-372). Recent work by Oden and others (see Oden 1978;
Oden and Spira 1978) can be brought to bear on the problem, at least as
regards literal and direct utterances. They have collected experimental
evidence supporting the idea that a semantic processor builds tentative
semantic structures in the course of language processing and applies
certain relevant semantic constraints to each of these structures to
determine their relative plausibility. This building and evaluation of
candidate semantic structures is to take place in parallel for all possi-
ble interpretations, just as proposed by the clausal-closure theory.
However, Oden's system includes two mechanisms that use "degree-
of-sensibleness" information to rule out interpretations, making it un-
necessary to process all of them completely:
~he first mech~ism, absolute judgment, rejects a candidate interpreta-
tIon whenever Its sensibleness value falls below some cut off point . . .
the second mechanism, relative judgment, eliminates all but the most
sensible interpretation, but only after enough processing has been per-
formed (e.g. at the end of a clause) so that the system has an accurate
assessment of the relative sensibleness of the interpretations. (Oden
1978, 35-36)
 Issues 256 Speech Act Schema and Psychology 257

S For us the main support for stage 3 comes from the results in Bever,
Garrett, and Hurtig (1973, experiment 2) which results support the
clausal hypothesis. Insofar as this model has empirical support, so does
NP VP our account of the steps from Ll to L2 of the SAS (see section 2.2),

I
S V
~NP whereby the hearer considers the various meanings of. the sentence
uttered and then rejects all but one as contextually inappropriate.

NP VP
/\
Det N
11.2.2. Propositional Content Stage

Disambiguation takes the hearer only part way to understanding what

A
Det N NP bothered
I
the
I
agent
is said. In addition, he must identify the speaker's referential intents.
As Fodor, Bever, and Garrett (1974, 142-170) illustrate, only recently
have psychologists given up the idea that all reference involves just
I resemblance or causation between referent and symbol, and seriously
N
considered the view that reference is in part a relationship involving
linguistic rules and conventions. As a result there is virtually no work
I
the Indians someone to report on the psychological reality of the inferences underlying the
Shoot}
{ shoot referential portion of the schema.
In chapter 8 we contrasted what is said with what is implied, en-
Figure 11.10 Structure underlying (14b) and one reading of (14a)
tailed, or presupposed. Psychologists have attempted in various ways
to test for psychological effects of these different factors. In particular,
S evidence has accumulated suggesting that some cases of pragmatic
presupposition are treated distinctly during comprehension and should
VP
be distinguished, psychologically, from entailments. Since we construe
NP
pragmatic presupposition in terms of what S presumes H to be aware
I
V
~NP of, we should examine the work on new and given information.
According to Clark and Haviland (1974, 1977), upon hearing a sen-
S

NP
~VP /\
Det N
tence like "Was it Margaret that Paul married?" where the given (pre-
sumed) information is that Paul married someone, H follows these
steps:
1

1
;

I
N
~NP
V bothered
I
the
I
agent
Given-New strategy:
GNI. Divide the sentence into presumed and new information.

~N
GN2. Match the presumed information in memory.
GN3. Integrate the new information with material now in memory.
Det
If this is a comprehension strategy, difficulties or complications at any
I I stage should increase appropriate measures of difficulty. Studies have
someone { shoot } the Indians used a variety of linguistic devices including the definite article, per-
shoot at sonal pronouns, Wh questions, repeating adverbs ("again"), relative
Figure 11.11 Structure underlying (14c) and one reading of (14a) clauses, and implicative verbs ("remember"). All tend to support the
 Issues 258 Speech Act Schema and Psychology 259

conclusion that steps GNI to GN3 are being followed during com- Table 11.2
prehension. Mean comprehension time (in milliseconds) for target sentences in direct,
Haviland and Clark (1974) report a sequence of experiments de- indirect, and negative antecedent pairs
signed, in part, to provide evidence for step GN2 of the strategy. In Adverbs
experiment 3, subjects were given sentences like (15)-(17): Antecedent still either again too means

(15) a. Last Christmas Eugene became absolutely smashed. Direct 1031 1102 984 976 1023
b. This Christmas he got very drunk again. Indirect 1058 1244 1040 1047 1097
(16) a. Last Christmas Eugene went to a lot of parties. Negative 1076 1141 1063 1065 1088
b. This Christmas he got very drunk again. Source: Haviland and Clark 1974.
(17) a. Last Christmas Eugene couldn't stay sober.
b. This Christmas he got very drunk again. Table 11.3
Problems and their mean latencies (in milliseconds) for experiment 1
In example (15) the context sentence (15a) provides an appropriate
Component Mean
antecedent for "again" in the second sentence (15b), and the match at Premise and Question Interrogated Latency
step GN2 should be quite direct. In example (16) the context sentence
John remembered to let the dog out.
(16a) provides only the basis for an inference to an appropriate match, Where is the dog? Implication 1795
so step GN2 would be less directly or immediately effected. In (17) the Where is the dog supposed to be? Presupposition 1939
context sentence specifies the appropriate condition negatively; an in- John forgot to let the dog out.
ference involving negation is required and thus is also less direct than Where is the dog? Implication 2199
(15). Subjects were timed from the beginning of reading the second Where is the dog supposed to be? Presupposition 2410
sentence to its being understood; mean latencies are reported in table It was thoughtful of John to let the dog out.
11.2. They confirm the plausibility of step GN2 of the strategy. Where is the dog? Presupposition 2015
Where is the dog supposed to be? Implication 2158
We mentioned in section 8.4 that some conversational implicatures,
I

unlike most, do not involve flouting a maxim and that H makes an It was thoughtless of John to let the dog out.
I I

Where is the dog? Presupposition 2441

inference on the assumption that S is obeying the conversational pre-
sumptions. It is plausible that step GN2 summarizes the mechanisms
underlying such an inference. Thus, in example (16), in order to main-
tain the presumption of relevance and apply step GN2, H must infer
that Eugene got drunk at a party last Christmas. In this case H applied
procedures of inductive inference; in other cases, he might use deduc-
tive inference or even have to restructure the utterance itself (see Clark
and Haviland 1977, 8).
Further evidence for such inferences comes from Brewer (1977). In
this study subjects heard a sentence such as "The safecracker put a
match to the fuse" and were given a cued recall test for these and
associated implications. Typically, more subjects recalled the prag-
matic implication of the sentence ("The safecracker lit the fuse") than
the original sentence itself.
In another experiment Just and Clark (1973) investigated pairs like
(18) and (19).
Where is the dog supposed to be? Implication 2426
Source: Just and Clark 1973.
(18) a. John remembered to let the dog out.
b. John forgot to let the dog out.
(19) a. It was thoughtful of John to let the dog out.
b. It was thoughtless of John to let the dog out.
In each pair the two sentences have the same (pragmatic) presupposi-
tion but opposite entailments. For example, both (18a) and (18b) pre-
suppose that the dog should be let out, but (18b) entails that it was not.
Just and Clark presented subjects with premise and question sets as in
table 11.3, and obtained the indicated latencies. Although "forget"
sentences take 438 msec longer than "remember" sentences, there was
no significant interaction effect by questions (1973, 24). The implica-
tions of "forget" and "thoughtless" did not take relatively longer to
Issues 260 Speech Act Schema and Psychology 261

answer than the questions interrogating the positive components. Just Table 11.4
and Clark concluded that subjects do not make use of presuppositions Problems and their mean latencies (in milliseconds) for experiment 2
and entailments independently of each other. Component Mean
In experiment 2, Just and Clark attempted to test the hypothesis (the Premise and Conclusions Interrogated Latency
ordered model) that subjects look for answers to questions first in the John remembered to let the dog out.
entailments of a sentence and then in the presupposition. Subjects were The dog is out. [true] Implication 2814
given premise-conclusion pairs ("If John remembered to let the dog The dog is in. [false] Implication 3252
The dog is supposed to be out. [true] Presupposition 3564
out, then the dog is out") drawn from table 11.4, and asked to judge The dog is supposed to be in. [false] Presupposition 4100
them true or false. The latencies obtained are indicated in the table. John forgot to let the dog out.
Assuming standard results on verification of positive and negative sen- The dog is in. [true] Implication 3670
tences, Just and Clark conclude that table 11.4 (and table 11.3 as well) The dog is out. [false] Implication 3536
provide support for the ordered-model hypothesis. The dog is supposed to be out. [true] Presupposition 4183
The dog is supposed to be in. [false] Presupposition 4664
If that hypothesis is correct, we might ask how it fits into the three
It was thoughtful of John to let the dog out.
steps of the given-new strategy. In that process, presumed material was The dog is out. [true] Presupposition 3647
matched and stored first and then entailments were processed; whereas The dog is in. [false] Presupposition 3964
in the ordered model, entailments are processed first, then presupposi- The dog is supposed to be out. [true] Implication 4162
tions. Is there an ordering paradox here? Not if one assumes that the The dog is supposed to be in. [false] Implication 4539
ordered model is in fact an elaboration beyond step GN3 of the given- It was thoughtless of John to let the dog out.
The dog is out. [true] Presupposition 3939
new strategy. That is, we should continue the strategy in such a way The dog is in. [false] Presupposition 4527
that when further access to the presumptions or implications of the The dog is supposed to be in. [true] Implication 4657
sentence is required, there is a step-wise procedure for carrying this The dog is supposed to be out. [false] Implication 4673
out: Source: Just and Clark 1973.

Given-New strategy continued:

GN4. Search new information for an answer to a question.
GN 5. If GN4 fails to be satisfactory, search the presumed information.
Additional evidence for inferential operations underlying communica-
tion at this level comes from the studies of Harris (1974), Singer (1976),
and Harris and Monaco (1978), though we will not review their results.
We conclude that insofar as such processing operations have psycho-
logical support, the sorts of inferences leading to L3 of the SAS can be
ruled plausible.
11.2.3. Illocutionary Force Stage
We tum now to the later stages of the comprehension process-iden-
tifying the speaker's illocutionary intent. For us the obvious initial
hypothesis concerning what underlies this identification is provided by
the information flow contained in the SAS, which can be summarized
as:
4. c. Direct act: Determine the literal meaning of e and then the literal
illocutionary act being performed, if any. If no literal act is being
performed, determine the nonliteral act being performed.
d. Indirect act: If S is not just performing a direct act (literal or
nonliteral), determine what indirect act is being performed as well.
e. Communicative intent: On the basis of 4(c) and 4(d) determine S's
communicative intent.
If this represents a psychologically real process, one can expect that (i)
H computes the literal meaning first and (ii) if H infers a nonliteral or
indirect intent, reaching that conclusion should take longer than com-
puting the literal meaning.
There is some evidence for inferences underlying the identification of
the direct force of an utterance, or at least its immediate storage in
memory. For instance Schweller, Brewer, and Dahl (1976) report the
results of two experiments on illocutionary force. In experiment 1,
subjects heard sentence triples such as the following:
Issues 262
(20) a. Implicit: The weatherman told the people about the approach-
ing tornado.
b. Control: The weatherman told the people about the approaching
warm weekend.
c. Explicit: The weatherman warned the people about the ap-
pro aching tornado.
Given the evidence that pragmatic implications and assertions can be
conflated in memory, it was predicted that there would be a shift in
recall of the (a) sentences in the direction of the explicit (c) sentences
but no such shift for the (b) sentences. The results (1976, 329) substan~
tiate the prediction with a 35 percent shift for the first type of sentence
and a 2 percent shift for the second type. Since inferences underlie the
pragmatic implications on which the prediction is based, we may ten-
tatively conclude that inferences underlie these misidentifications of
illocutionary force on recall, and so are available to play a role in
comprehension. For more direct evidence we must look at nonliteral
and indirect acts.
An experiment by Brewer, Harris, and Brewer suggests that predic-
tions (i) and (ii) are correct for nonliteral utterances. In their experi-
ment subjects were given pairs of sentences from proverb sets such as:
Original Proverb: It's a silly fish that is caught twice with the same bait.
Literal-Same paraphrase: Only a foolish fish is captured more than once
on the same hook.
Figurative-Same paraphrase: Only a fool does not learn from experi-
ence.
Literal-Different paraphrase: A wise fish and a foolish fish, both caught,
are equally dead.
Figurative-Different paraphrase: Wise men as much as fools do not
learn from experience.
The reasoning was that if a hearer heard a literal paraphrase-plus-
proverb sequence then there should be little discernible difference
between this order and the reverse order of proverb-plus-literal para-
phrase. However, for the figurative paraphrase-plus-proverb the hearer
will first have to calculate the literal meaning of the proverb and then its
figurative meaning before being able to match it with its correct para-
phrase. This extra step should consume extra time, so there should be,
and was, a significant interaction between order of proverb presenta-
Speech Act Schema and Psychology 263
tion and (speed of) recognition of literal versus figurative paraphrase.
The authors concluded that the processing of literal and figurative
meaning occurs in two stages and that the comprehension of the literal
meaning precedes the comprehension of the figurative meaning.
Because only proverbs unfamiliar to the subjects were used, we
interpret the experiment as providing evidence for the psychological
reality of the presumption of literalness (PL). In effect, subjects were
first reacting to the proverbs as if they were literal, then processing
them figuratively when the task demanded it. Under the circumstances
that means that they were presuming that the sentence should be taken
literally until the task proved that assumption wrong, and this is tanta-
mount to conforming to PL. It might be interesting to compare these
results with similar tests involving familiar proverbs, which would
probably override the PL and so not add significantly to response
times.
Clark and Lucy (1975), using the familiar three-feature pattern for
negative and positive sentences, found that at least sometimes predic-
tions (i) and (ii) are true for indirect acts as well. The three characteris-
tics of this pattern are that, other things being equal, (1) positives are
judged true faster than false, (2) negatives are judged false faster than
true, and (3) positives are easier to process, overall, than negatives.
Subjects were presented with displays consisting of a sentence like
(21a) or (21b) followed by a circle colored either pink or blue (sentence
pairs used are listed in table 11.5).
(21) a. Please color the circle blue. (positive indirect force)
b. Please don't color the circle blue. (negative indirect force)
Subjects were to respond "yes" (true) or "no" (false) by pushing an
appropriate button, on condition that the displayed circle fulfilled the
directive conveyed by the sentence. Since each pair of sentences con-
tained a negative element, it was possible to see whether the three-
feature pattern emerged with respect to the indirect force of the
sentence. The pattern did emerge (p. 62):
True: average of 346 msec faster for positive requests.
False: average of 308 msec faster for negative requests.
Overall, positives average 222 msec faster than negatives.
These results, as well as analysis of the individual pairs of sentences
given in table 11.5 indicate that subjects did represent the indirect force
of the sentence in the course of comprehension.
Issues 264 Speech Act Schema and Psychology 265

Table 11.5 This experiment provides evidence for predictions (i) and (ii). Evi-
Mean latencies and percent errors a for pairs I-lOb dence that H computes literal meaning first comes, for example, from
Response sentence pairs 9 and 10 (table 11.5). On the assumption that "unless" is
Pairs Basic sentences True False Mean inherently negative (= "if not"), it should take longer to encode "un-
1. (a) Please color the circle blue. 1213 (0) 1610 (5) 1411
less" than to encode "if," and it did take over Y2 sec longer to verify
(b) Please don't color the circle blue. 1799 (10) 1644 (12) 1722 pair 10 than pair 9. This suggests that literal meaning was computed in
2. (a) Can you make the circle blue? 1473 (0) 1990 (1) 1731 the course of carrying out the verification task, which of course took
(b) Must you make the circle blue? 2082 (16) 1810 (2) 1946 longer. Moreover, this time difference in verification cannot be attrib-
3. (a) Why not color the circle blue? 1510 (1) 2060 (5) 1785 uted to a difference in indirect force, since it is plausible to assume that
(b) Why color the circle blue? 2047 (11) 1856 (6) 1951 verification of corresponding members of pairs 9 and 10 is task-
4. (a) I would love to see the circle 1537 (0) 1771 (0) 1654 equivalent:
colored blue.
(b) I would hate to see the circle 2014 (5) 1778 (1) 1896 (a) I'll be very happy if ... = I'll be very sad unless . ..
colored blue. (b) I'll be very sad if . . . = I'll be very happy unless . . .
5. (a) You should color the circle blue. 1613 (11) 1662 (3) 1637
(b) You shouldn't color the circle blue. 1978 (3) 1669 (8) 1824 Further evidence that literal meaning is computed is that interrogatives
6. (a) Shouldn't you color the circle 1723 (2) 2047 (2) 1885
consistently took longer than their corresponding dec1aratives. It is
blue? difficult to attribute this always to differences in the length of the sen-
(b) Should you color the circle blue? 2510 (16) 1945 (16) 2228 tences or differences of indirect force (Clark and Lucy 1975, 66ff).
7. (a) The circle really needs to be 1544 (3) 1916 (5) 1730 Although this study supports the SAS in broad outline, it leaves most
painted blue. details open. For instance, by instructing subjects at the outset to look
(b) The circle doesn't really need to be 2156 (5) 2122 (2) 2139
painted blue.
for directive force, most of the inferential apparatus of the SAS was
by-passed. After all, the SAS comprises numerous lines of inference,
8. (a) Doesn't the circle really need to be 2098 (5) 2373 (7) 2236
painted blue? with citations of a variety of presumptions and mutual contextual be-
(b) Does the circle really need to be 2251 (9) 2087 (6) 2169 liefs. The Clark and Lucy study taps only two of these: the literal
painted blue? meaning of e and the indirect force of its utterance. It even leaves open
9. (a) I'll be very happy if you make the 1779 (0) 2103 (3) 1941 the question of whether or not H infers that S has performed a literal
circle blue. illocutionary act.
(b) I'll be very sad if you make the 2362 (7) 1880 (2) 2122
circle blue. We have not yet said anything about the perlocutionary act from a
psychological point of view. Schweller, Brewer, and Dahl (1976) report
10. (a) I'll be very sad unless you make 2357 (2) 2798 (10) 2577
the circle blue. results of experiments on perlocutionary acts. In one experiment (ex-
(b) I'll be very happy unless you make 2692 (11) 2322 (8) 2507 periment 4) subjects were presented with sentences describing an illo-
the circle blue. cutionary act such as (22a).
a In parentheses.
b Latencies are given in msec. (22) a. Illocution: The angry farmer threatened the trespassing boys.
Source: Clark and Lucy 1975. b. Appropriate perlocution: The farmer frightened the trespassing
boys.
c. Inappropriate perlocution: The farmer calmed the trespassing
boys.
Asked later to recall what they had seen, many subjects who had in fact
seen an illocutionary description such as (22a) instead reported the
Issues 266

appropriate perlocution (22b) but not, of course, the inappropriate Appendix Mutual Belief and Social
per!o~ution. (22c). Subjects conflated in memory the illocutionary de- Concepts
scnptlOn wIth the appropriate perlocutionary description, and this sug-
gests that there is a readiness to infer from the original illocution what
we call (in chapter 3) the associated perlocution. Schweller, Brewer
and Dahl conclude: '
Th~ overall results of these four experiments demonstrate the highly
act~ve nature ~f a hearer's processing of sentences and the effects this
actIve proceSSIng has on memory for sentences. Subjects are clearly
~ble to use ,:onten! and context to elaborate sentences they have heard
In terms of IllocutlOnary forces and perlocutionary effects . . . earlier
work . . . has shown that a number of different kinds of inferences can
become confused in memory with the material originally presented. The notion of mutual belief has figured prominently in our account of
The results .of the present experiments extend this range of inference illocutionary acts. There are mutual contextual beliefs, which facilitate
types affectIng memory to include illocutionary forces and perlocution-
ary effects, thus lending further empirical support to the general theory various steps of the hearer's inference to the speaker's communicative
tha~ human. mem~~)fy is not a passive, isolated system, but rather an intention, and the several presumptions, which assure the hearer that
actIve o~e In ~hIch ~nowledge from one domain interacts in a very there is an inference to be drawn. The speaker relies on these mutual
systematIc fashIon wIth knowledge from other domains. (1976, 336) beliefs to make his communicative intention recognizable. The notion
of mutual belief was central also to our discussion of conventional
Although we have been able to survey only briefly some of the rele-
illocutionary acts. The concept of convention was explicated in terms
vant literatQre, the limited evidence currently available supports sev-
of mutual belief, and the institutional facts that conventional acts affect
eral steps of the speech act schema. In particular, we found evidence
or effect are matters of mutual belief.
for contextual selection of one reading of an expression from among
We wish to broaden our perspective and explain the systematic role
possible readings, even when subjects were not aware of the ambiguity.
of mutual belief in the analysis of various sociological concepts. These
We also found evidence for representation of the nonliteral and indirect
concepts enable us to represent a variety of institutional facts and
force of an utterance, as well as for inferential operations connecting
phenomena of social interaction, of which communication is a special
them to the literal and direct force. Finally, we noted some evidence
case. We will be concerned with two kinds of sociological concepts,
against total decomposition of lexical items during comprehension,
social regularities and social collectivities. Among regularities we will
evidence thus favoring the inferential semantics tentatively endorsed in
distinguish norms, practices, rules, and roles, and these notions will be
chapter 8. Considered as a whole, this evidence is encouraging for the
subdivided further. Social collectivities can be distinguished by degree
inferential approach to linguistic communication, but it would be rash
of structure: types, groups, and organizations.
at this time to claim more than preliminary empirical support for our
The underlying point of using the notion of mutual belief to analyze
theory. The SAS represents much more information than current ex-
sociological concepts is to resist reifying society. Society is not an
perimental literature touches on, and it encompasses a variety of in-
autonomous entity to which people are subject, even if they often
ference patterns. Research on the psychological implications of the
experience it that way, so we must be careful not to abstract it from the
speech act schema has only just begun.
people who make it up. On the other hand, society does not exist in the
privacy of people's heads, as wishes and worries do, any more than it
exists "out there," as tigers or trees do. In some sense it is intersubjec-
tive, and the concept of mutual belief enables us to explain precisely
how. We can think of society as a system regulating and organizing
 Appendix 268
people's behavior. Our strategy will be to analyze the concepts in terms
of which this regulation and organization can be represented and to do
so in terms of mutual belief. There would be no social system if people
had nothing to do with one another, but more than that, their actions
interlock and exhibit patterns in virtue of the beliefs and attitudes they
share. In fact, a great part of what people know about one another is
that they share a great many beliefs and attitudes. Knowing this re-
quires a shared conceptual scheme, which, besides its elaborate cate-
gorization of things in the natural world, includes a catalog of what
there is for people to be and to do. When we refer to social regulation
and organization, we imply that the resulting patterns of behavior are
not merely statistical, meeting the expectations of detached scientific
observers, but are in large measure socially recognized, meeting the
expectations of members of society themselves. At least in part, the
system of regulation and organization that explains these patterns is
internalized in people's beliefs and attitudes and is part of their shared
conceptual scheme.
The sociologists Klapp (1957) and Scheff (1967) first put forth the
provocative philosophical suggestion that the basic concepts of sociol-
ogy can be derived from what they called the notion of consensus. They
and other writers (many cited by Scheff) have thought this notion
relevant to such diverse topics as public opinion, mass action, norms
and roles, communication, games, culture and tradition, socialization,
and social cohesion. Use of the concept of consensus is widespread in
!' social science, but little effort has been made to formulate that concept
as anything more than individual agreement. Scheff cites numerous
examples of experimental and theoretical work in which this simplistic
conception is implicit. The trouble with viewing consensus simply as
agreement, Scheff points out, is the failure to allow for the possibility of
pluralistic ignorance (people agree but don't realize it) and false con-
sensus (people mistakenly believe that they agree). Pluralistic igno-
rance satisfies the simple definition but lacks the behavioral effects of
genuine consensus, whereas false consensus may have the same effect
on behavior as genuine consensus even though agreement is lacking.
Various turns of phrase have been used, Scheff notes, to take these
phenomena into account-perceived consensus, the generalized other,
interpenetration of perspectives, and reciprocity of perspectives - but
these do not constitute an explication of consensus.
Scheff proposes an analysis in terms of levels of agreement. Among a
group of people with opinions on a given issue, a majority can agree or
d
Mutual Belief and Social Concepts 269
fail to agree at the first level, that is, on the issue itself. At the second
level they could understand or fail to understand that they agree (or
disagree); pluralistic ignorance is a lack of understanding about agree-
ment, and false consensus is misunderstanding about disagreement. At
the third level there can be realization or lack of realization about
understanding (or lack of it) about agreement (or disagreement). Theo-
retically, further levels could be brought in, but practically, on Scheff's
analysis, consensus consists in majority agreement, understanding, and
realization. Assuming the majority to be a determinate part of the
whole group and taking three levels into account, Scheff notes that
consensus is but one of eight possible situations.
The trouble with Scheff's account of consensus is its implicit as-
sumption that there is a determinate majority, which agrees or dis-
agrees at any of the three levels. This difficulty can be avoided and
consensus defined more perspicuously, we suggest, in terms of belief
rather than agreement. For us consensus in the sense important to
social scientists is mutual belief. In our formulation (as well as in our
later analyses), the usual if and only if of philosophical definitions will
be supplanted by to the degree to which, represented for convenience
by ex:, the mathematical symbol for proportionality. In this way we
accommodate the fact that the phenomena under consideration admit
of degree and that no clear line is to be drawn between cases where the
concept in question does apply and cases where it does not. We recog-
nize, however, that since multiple dimensions are involved, a metric
needs to be specified. At any rate, we define mutual belief (over a
collectivity G with respect to a proposition p):
DF 1: It is mutually believed in G that p ex: the members of G believe:
i. that p,
ii. that the members of G believe that p, and
iii. that the members of G believe that the members of G believe that
p.l
This definition clearly captures such ideas as the interpenetration or
reciprocity of perspectives. And without itself reifying society, it sug-
gests how people can reify society when, in Laing's words (1968, ch. 4),
they conform "to a presence that is everywhere elsewhere": the mem-
bers of G may think of the members of G as society rather than as
everyone else, and those who think this may include everyone.
This notion of mutual belief figures centrally in our definitions of
various concepts of social regularities and of social collectivities; their
Appendix 270
definability in terms of mutual belief is meant to demonstrate and expli-
cate the intersubjective character of these concepts. We often speak of
something's being psychologically real to someone if he believes it
exists, if it exists for him, irrespective of whether it really does exist.
Similarly, we may use the phrase socially real to apply to anything that
exists for a given society or collectivity G. However, for something to
be socially or intersubjectively real for G, it must be not merely psy-
chologically real for each member of G. After all, each might fail to
realize this and perhaps even believe that it exists for him alone. Even
if that is not the case and the members of G believe it to be psychologi-
cally real for one another, they might not believe or might even dis-
believe that others believe that. To be socially real, something must be
mutually believed to exist. This is the status of the regularities and
collectivities to be defined.
Our employment of the notion of mutual belief should not give the
impression that people's second- and third-level beliefs are automati-
cally uniform and stable. It is an empirical question, beyond the scope
of philosophical analysis, how people acquire the beliefs they do and
why there is the uniformity there is in a given culture. Many topics in
social psychology bear on this question: socialization, conformity and
deviance, mass communication, and the social self. Moreover, mutual
beliefs don't come into existence all at once; they develop over a period
of time. In the United States mutual belief in the wrongness of Ameri-
can involvement in the Vietnamese war took years to be realized.
Indeed, the Administration's stratagem of singling out the so-called
silent majority created a long period of false consensus. Fads, fashions,
and crazes seem to involve flowing and ebbing waves of mutual belief.
Such phenomena as social movements, political backlashes, trends in
the stock market (panic selling), landslide elections resulting from the
bandwagon effect, keeping up with the Joneses, generation gaps, and
religious revivals all seem to involve upsurges of mutual belief. There is
no telling the degree to which mass media and the astute exploitation
thereof may foster a state of pluralistic ignorance about the need for
change and, further, a state of false consensus concerning the status
quo.
SOCIAL REGULARITIES
Three kinds of social regularities may be distinguished: norms, prac-
tices, and rules. The reasons for distinguishing them will be ex-
plained-in fact they are embodied in the definitions we will offer-
Mutual Belief and Social Concepts 271
and subconcepts and kindred concepts of each will be singled out. The
key notion of mutual belief figures in each definition.
Norms
What is a social norm? Landis (1971, 228) defines a norm as "the
accepted or required behavior for a person in a particular situation."
According to Secord and Backman (1974, 300), "a social norm is an
expectation shared by group members which specifies behavior that is
appropriate for a given situation." They note (p. 402) that expectations
are both anticipatory and normative in nature, and thus social norms
are also. This descriptive/evaluative ambiguity is suggested by the ad-
jectives "normal" and "normative," so a norm is both a regularity and
a regulation. Landis's definition mentions merely the normative aspect
of norms; but it seems that to count as a norm, the behavior in question
must be generally performed and not merely a generally unfulfilled
standard. However, we will follow Landis in calling the expected be-
havior rather than the expectation itself the norm. Nothing crucial rides
on this terminological preference, but it will facilitate the formulation
of our definition.
Our definition must embody the idea that a social norm is neither
merely a statistical regularity nor merely a standard that people may
hardly ever observe; thus, the definition must exclude behavior like
people's putting on their pants before their shoes and standards like
never showing temper. Our definition must express the social reality of
norms. We letA designate the kind of behavior in question,2 C the kind
of recurrent situation to which the norm applies, and MB 'mutually
believed.'
DF 2: A (in C) is a social norm in G ex::
i. the members of G do A in C,
ii. it is MB-ed in G that (i), and
iii. it is MB-ed in G that the members of G should do A in C.
It follows from this definition that when C arises, the mutual beliefs
specified will be activated in the people involved, and so, assuming
they identify one another as members of G, they will mutually expect
one another to do A. In specific instances the specification of A may
designate only a certain kind of agent, such as children, drivers, or
guests. In this case only the specified kind of agent can fulfill or violate
the norm, but still the mutual beliefs are shared by all.
Among the types of action that the variable A ranges over are nega-
Appendix 272
tive actions, such as not picking your nose or not embracing strangers.
When such a description is plugged into the definition, what the mem-
bers of G do is not to perform a certain action. For such negative
norms, as we might call them, we may formulate a separate definition.
DF3: There is a social norm in G against A (in C) 0::
i. the members of G do not do A in C,
ii. it is MB-ed in G that (i), and
iii. it is MB-ed in G that the members of G should not do A in C.
Logically speaking, this definition is superfluous, provided we allow A
in DF2 to range over negative actions. Nevertheless, it is worth for-
mulating this definition of negative social norms, for a great many
social norms are of this type, proscriptions rather than prescriptions.
(For subsequent definitions the negative forms will be omitted.)
It is possible, indeed common, for just some of the clauses in DF2 to
be satisfied. Several cases are worth labeling:
DF4 : A (in C) is a statistical regularity in G 0::
i. the members of G do A in C.
Sometimes the term norm (as opposed to social norm) is used to desig-
nate statistical regularities.
Statistical regularities need not be recognized by the people to whom
they apply, and social scientists have discovered a great many such
regularities (including many pertaining to topics other than behavior).
We should single out the case where the members of G themselves are
aware, indeed mutually aware, of such a regularity, for in this event,
when C arises there will be a mutual expectation that A be done.
DFs: A (in C) is a social regularity3 in G 0::
i. the members of G do A in C, and
ii. it is MB-ed in G that (i).
Since clause (iii) ofDF2 is absent from DFs , the mutual expectation that
arises in C need not be normative in nature but merely empirical.
Notice that clause (ii) might hold without (i) holding, in which case the
MB would be false. Such a socially imagined regularity might have
occurred-such a false mutual belief might have prevailed-as to sex-
ual behavior when it was a hush-hush subject.
Finally, we might define the case where a certain type of behavior is
mutually recognized as the standard even when people generally don't
live up to it.
Mutual Belief and Social Concepts 273
DF6: A (in C) is a social standard in G 0::
iii. it is MB-ed in G that the members of G should do A in C.
Of course a standard may also be a statistical regularity, and if it is a
social regularity as well, then it is a social norm. A social standard is a
special case of a social value, which might be defined as anything (not
just behavior) that is mutually believed in G to be of value.
Behavior that is the standard, including that which is the norm, is
mutually believed in G to be what people should do-or should not do,
in the case of negative norms. In between is the entire range of things
that are socially acceptable, neither required nor forbidden. Accord-
ingly, we can say that a type of behavior is socially acceptable if there
is no standard that requires not doing it, that is, no standard against it.
Having defined some of the concepts that fall under it, let us return
briefly to the concept of social norm itself, as analyzed in DF2. Even
though the definition implies that when C arises people mutually expect
A to be performed, it implies nothing about people's reasons for doing
A. A person utilizing the expectations implied by the definition might
do A to receive approval, to avoid being socially rejected (or punished),
to appear mentally normal, to avoid suspicion. He may have no reason
at all and do A out of habit. Why people conform to norms depends on
the norm and on the person, but this is a topic for social psychology not
philosophical analysis.
DF2 does not imply that there must be a sanction for every norm-
some norms are more serious than others, and some people are more
serious than others about norms. Nevertheless, because norms are
mutually expected to be followed, it is not surprising that when some-
one violates a norm, other people may react accordingly. As Hart
(1961, 54-56) puts it, norms have an "internal aspect," in that they
provide a basis for "criticism of others and demands for conformity"
that should yield" acknowledgement of the legitimacy of such criticism
and demands when received from others." In short, norms are not only
followed but applied. Indeed, there are norms regarding the violation of
norms. Depending on the seriousness of the offense, there may be a
norm for others to criticize the violator and to demand amends or at
least apology. But in some cases there is a norm against saying any-
thing-it would be impolite, cause embarrassment, or create a scene.
And of course, again depending on the case, there are norms for vio-
lators: to apologize, to offer a legitimate excuse, and where possible to
make amends. Still again, there is sometimes a norm against the of-
fender saying anything.
Appendix 274
Practices
That a type of act is the norm does not determine what people's reasons
are for doing it. Even if a person believes he should do it, that may not
be his reason for doing it. However, a special case of social norms,
which we label social practices, has a reason for action built into its
definition. There is a definitional connection between the existence of a
practice and people's reason for following it: a practice is a norm that
people follow in order to conform, in order to be like others or at least
to seem to be. Norms in general can be followed for this reason, but
practices by definition are.
DF7: A (in C) is a social practice in Go:::
i. A (in C) is a social norm in G, and
ii. people's reason for doing A in C is, at least partly, that it is generally
done.
The definition of norms (DF2) implies that people who follow a norm
recognize that others do, but it does not imply that this is their reason
for doing so, even if part of their reason is that following the norm must
be the right thing to do because others do it-that's not the reason
specified in clause (ii) of DF7. Typical examples of practices include
dressing in accordance to one's sex (or class or age), social drinking,
and standing for the national anthem.
It might be objected that a vicious circle is built into the definition of
practices: the members of G do A partly because the members of G do
A. However, this is not a vicious circle but an endless chain. Each does
A partly because others have been doing A (and can be expected to
continue to). Naturally, that could not have always been everyone's
reason, but the definition does not require that A was originally a
practice. During the period in which a practice comes into being, as
people begin performing the action in question, their reasons for doing
it and their beliefs about the extent of its being done may be unclear and
in flux, whereas once the practice is established and stable, mutual
belief in its existence and continuation prevails.
Rapidly changing practices, such as fads and fashions, are an in-
teresting special case. Dance crazes, hair styles, and dress fashions are
practices that people follow in order to do what is "in," to be "with it"
(these locutions are no longer in fashion). Most people don't want to
appear "out of it," and as styles periodically change, the idea is to be
neither too far ahead of the game nor too far behind it (style setters are
Mutual Belief and Social Concepts 275
a special type). Notice, by the way, that the collectivity G in which a
fad or fashion prevails maf be specifiable only as that in which the fad
or fashion prevails- the in crowd consists of those who do what is in.
Rules
The term social rule will be used to designate any social norm that
entails some special social obligation (in addition to the general obliga-
tion to follow norms).4 Four different kinds of rules warrant being
defined here. Since there is no precedent (to our knowledge) for ter-
minology that marks these distinct kinds, we call them, for reasons that
will become evident, coordinative rules, cooperative rules, collective
rules, and regulations. These categories are not mutually exclusive.
Coordinative rules are social rules like driving on the right, speaking
in tum, and going to the end of the line, whose point is to coordinate the
activity of a number of people who are trying to do more or less the
same thing with a minimum of interference from others. Each person
who follows such a rule is doing his part in a joint effort to coordinate
the activities (driving, communicating, buying a ticket) of all involved.
DFs: A (in C) is a coordinative rule in G 0:::
i. A (in C) is a social norm in G, and
ii. it is MB-ed in G that general performance of A in C enables the
members of G each to do some act B with a minimum of interference
from one another.
The definition requires only that general performance of A be mutually
believed to minimize interference-whether it does so is another
matter.
Cooperative rules. There are many actions whose performance by
one person makes sense only if many others do the same thing: re-
cycling cans and bottles, conserving energy, going out on strike. If one
has good reason to believe that hardly anyone else will do A, one has
little reason to do it oneself. These cumulative actions (see Bach 1977)
yield collective goods or prevent collective evils. Collective goods are
goods that everyone enjoys if anyone does, such as clean air, quiet in a
library, or higher wages (in a union shop). Collective evils hurt every-
one (not that a few don't benefit from them also) if they hurt anyone, as
with air pollution and oil shortages. The free-rider problem arises when
people refuse to perform a cumulative action (conserving energy or not
littering) because enough others will do it anyway. The problem is that

Appendix 276
if too many people think this way, too few will perform the cumulative
action-unless there is a rule requiring each person to contribute his
effort. Such a rule is a cooperative rule.
DF9: A cumulative action A is a cooperative rule in G ex::
i. A is a social norm in G, and
ii. it is MB-ed in G that general performance of A tends to produce
some collective good or prevent some collective evil.
Clause (ii) requires only a mutual belief not the truth of it, so conceiva-
bly general observance of a cooperative rule could turn out to be detri-
mental to G.
Collective rules organize collective action out of individual efforts.
Rules that divide labor in a family, a tribe, or a company are collective
in this sense. In general, a collective rule apportions assignments to
different members of G, whose combined action is presumed to further
the interests of G or its members. Thus, to follow a collective rule is to
do one's part.
DF to : A set of actions At, ... ,An is a collective rule in G ex::
i. to do one's part is the social norm in G,
ii. there is a mutually recognized procedure for determining each per-
son's part, and
iii. each person's part is one of At,. . . ,An.
The definition does not require each member of G to be aware of what
everyone else's assignment is but only that everybody has one (several
might have the same). In a formal organization (to be defined later),
there are people whose part is to determine others' parts, whereas in an
informal group the apportionment of parts could be arrived at by
mutual agreement rather than executive edict. We will see that a formal
organization can be thought of as governed by a system of collective
rules.
Regulations are standards that are enacted, promulgated, or other-
wise imposed on G by some mutually recognized authority (who may
or may not be legitimate- having power does not imply having the
right to exercise it). Generally a regulation is codified or otherwise
made official, but this is not required by our definition.
DF 11: A (in C) is a regulation in G ex::
i. A (in C) is a social standard in G, and
ii. by virtue of some mutually recognized authority, A (in C) is required
in G.
Mutual Belief and Social Concepts 277
Political regulations are laws, and regulations generally (but not neces-
sarily) have sanctions enforcing their observance, but the definition
requires only that they be standards, not norms. Nevertheless a regula-
tion can be a norm-or even a coordinative, a cooperative, or a collec-
tive rule.
SOCIAL COLLECTIVITIES
People can be grouped either in the sense of being categorized together
or in the sense of interacting with one another. Sociologists generally
call the first sort of grouping a category, reserving the term group for
the second. The only relationship members of a category have, gener-
ally, is that of having something in common, the feature that places
them in that category; the relationship among members of a group is
their pattern of interaction. Further distinctions must be drawn here,
and they will be the subject of our detailed analysis. A role category is a
socially recognized category whose behavior is subject to specific ex-
pectations. Role categories include both positions and social types. The
notion of role itself must be explained; there are ambiguities in the
notion of role (akin to those in the notion of social norm) that have to be
resolved. Under the heading of groups we will distinguish, in increasing
order of structure and complexity, face-to-face groups, crowds, social
groups, and organizations. The notion of mutual belief will have a
central place in all our definitions.
Role categories
People can be classified in all sorts of ways: by age, sex, race, height,
personality, nationality, profession. Some categories are not recog-
nized by the general public but only by social scientists, and of course
among them there is divergence, for .instance, on the categorization
of personality types'. Not every category is worth singling out, either
because it is nearly vacuous (blue-eyed teen-aged businesswomen)
or because no significant empirical generalization can be made over
it (left-handed pawnbrokers born on odd-numbered Thursdays). We
should distinguish categories that are socially recognized, and over
which there are social expectations, from those recognized only by
social scientists and over which there are merely statistical regularities.
We will follow Secord and Backman (1974,402) and use the term role
category. Sociologists commonly use status to designate a position in a
group or in a social structure. People also classify one another by
Appendix 278 Mutual Belief and Social Concepts 279

personality traits and behavioral idiosyncrasies; these social types in- as social regularities, role ~tandards as social standards, and normative
clude fools, egotists, heroes, nice guys, creeps, bigots, big shots, and roles as combinations of both. Thus we let R range over patterns of
cheapskates. We will use the term position, rather than status with its behavior (since roles typically include a cluster rather than a single kind
evaluative connotations, to designate a place in a group or social rela- of behavior).
tion; role category will cover both positions and social types, depending
on the features (F) involved. DF13: R is a role for people of role category F in G 0::
i. in G people with F exhibit R, and
DF 12 : A set of people with F is a role category in G 0:: ii. it is MB-ed in G that (i).
i. it is MB-ed in G that F is used to classify people in G, and DF 14 : R is a role standard for Fin G 0::
ii. there is some social regularity, standard, or norm applying to people iii. it is MB-ed in G that in G people with F should exhibit R.
with F.
DF 1S : R is a normative role for F in G 0::
Clause (i) says that it is mutually recognized in G that people classify i. in G people with F exhibit R,
one another in terms of F and not-F (American/foreign) or along a ii. it is MB-ed in G that (i), and
dimension on which F lies (Italian, on the dimension of nationality). iii. it is MB-ed in G that in G people with F should exhibit R.
Clause (ii) requires that F not be some idle category that doesn't pertain
Generally speaking, roles but not role standards apply to social
to the behavior of people with F and to people's mutual beliefs about
types, like fools, bigots, and cheapskates; whereas positions, like par-
the behavior of people with F. In this way, role categories are relevant
ents, preachers, and politicians, have role standards (or even normative
to people's social behavior and awareness of it, for they apply to people
roles, if it is mutually believed that they meet their standards and they
about whom there are special mutual expectations. 5 In discussing roles,
do). It is important to make the distinctions embodied in these defini-
we will explain why clause (ii) requires either a social regularity, a
tions, because the expectations (mutual beliefs activated in particular
standard, or a norm to apply to people with the features defining a role
situations) that are directed at people of a given type or position can be
category.6
anticipatory, normative, or both. Which they are depends on the type
The term social role (or simply role) is highly ambiguous. Secord and
of position and on people's associated mutual beliefs.
Backman (1974, 405) note that it can refer to a role category, to the
The reason social types are generally associated with roles and posi-
expectations associated with that category (role expectations), or to the
tions with role standards is that roles generally define social types
expected behavior itself. We will restrict the term role to the expected
whereas positions generally define role standards. In other words, a
behavior, although this is not to be confused with what Secord and
person belongs to a type (fool, hot shot, ladies' man) in virtue of what
Backman (pp. 405-406) call "role behavior," which is whatever a
he does; a person does certain things because of his position (coach,
person in a given role category actually does relevant to the role ex-
custodian, columnist). However, there are exceptions to this generali-
pectations, whether or not he fulfills them.
zation. Social types like nudists and cigar smokers are subject to role
If this isn't ambiguity enough, restricting the term role to the ex-
standards; other social types, like ladies and heroes, are defined by the
pected behavior must allow for the fact that expectations can be nor-
standards they live up to (normative roles).
mative or merely anticipatory (as we observed in discussing norms).
Roles (and role standards) often interlock in the sense that they may
Our account of the notion of role should reflect this distinction, since
specify how people of one position (or type) act (or are to act) with
some roles are expected (anticipated) but hardly required to be played.
regard to people of the same or of some other position (or type). Or-
Everybody expects a boor to act rudely, but nobody (besides himself)
ganizations (DF 17) are a complex instance. A final point about our
thinks he should. Accordingly, the use of the term role must be re-
definitions is that when clause (ii) is satisfied without clause (i), we have
stricted further and other terms introduced to fully resolve the am-
the case of a social stereotype, where there is a (usually) false mutual
biguities inherent in standard usage. We will distinguish between roles
belief in G about some minority's behavior.
Appendix 280
Groups
In the broadest sense of the term, a group can be a category, an aggre-
gate, or a role category, but here, following general usage in social
science,1 the term group will be more restrictive. Even so, groups can
range from strangers in brief, perfunctory interaction to a formal or-
ganization. Despite the variation on this continuum, groups all have the
feature that their members have some sort of structured relationship,
be it mutual expectations based on social norms that are applied anon-
ymously or a rich set of mutual beliefs based on personal acquaintance.
We distinguish four kinds of groups: face-to-face groups, crowds, social
groups, and organizations.
No special definition (in terms of mutual belief) is needed for a
face-to-face group, which is any combination of people engaged in any
kind of direct interaction, however brief, casual, and routine. The per-
sons involved may be acquaintances or members of the same social
group or organization, in which case specific norms and role standards
apply. But even unacquainted and unrelated participants in the most
fleeting random encounter have mutual expectations based on general
social norms and on contextual mutual beliefs as to their respective
identities and the definition of the situation.
Not just any aggregate of people in proximity constitutes a group.
Pedestrians on a busy mall, passengers on a commuter train, and shop-
pers in a department store do not constitute a group- there need not be
any interaction or structured relationship among co-present persons
over and above that following from the observance of general social
norms, such as common courtesies. There is no action as a group or as
members of a group. However, in the event of a sudden emergency or
anything which captures everyone's attention, collective action quickly
becomes possible. Suddenly they become aware-indeed mutually
aware-of the same thing. Then they become a crowd, which Landis
(1971, 211) defines as 'a temporary collection of people in close physi-
cal contact reacting together to a common stimulus.' We add the re-
quirement that members of a crowd be mutually aware of the stimulus;
the phrase "reacting together" does not imply this awareness-they
might be reacting simultaneously but without being mutually aware of
the stimulus or of each other. Tht(re are many kinds of crowds, includ-
ing audiences, mobs, and social gatherings.
People having something in common need not have any common
Mutual Belief and Social Concepts 281
interest or mutual awareness. Th,ey need not have any special pattern
of interaction. What we call a social group has these features. This
accords fairly well with the common sociological idea of a social group
as entailing consciousness of kind (or a we feeling) together with pat-
terned interaction. Our definition is meant to explicate this idea.
DF 16 : G is a social group ex:
i. the members of G have some feature F in common, together with an
associated interest [,
ii. it is MB-ed in G that (i),
iii. partly because of (i) and (ii), the members of G think of G as "we,"
and
iv. partly because of (i), (ii), and (iii), there is a pattern of interaction in
G governed by a set of norms (including cooperative rules) and norma-
tive roles, MB-ed in G to further [.
Social groups range from friendships, families, and cliques to clubs,
unions, and professional associations (the latter groups have organiza-
tions within them). DF 16 is intended to capture the idea of a collection
of people mutually regarding themselves as a unit (we) to which they
belong, united by having something in common, including a mutual
interest that is presumably furthered in their group activities. All four
clauses are necessary because, for example, clauses (i) and (ii) can be
satisfied without (iii) or (iv), as in the case of social types like baseball
fans or fat people. Indeed, the first three clauses can be satisfied with-
out the fourth, as illustrated by an oppressed class (or race) with what
Marxists call class consciousness but with no organized activity. A
social group, at least in the restricted sense defined here, must have the
features specified by all four clauses.
The mutual beliefs associated with social types and positions (see
DF 12) prevail in the community at large, whereas those associated with
social groups need not extend beyond the group itself. In other words,
membership in a social group, as opposed to a role category, is deter-
mined by the members themselves. To become a member of an existing
social group, a person must be regarded by the established members as
a member; to be so regarded he may have to fulfill some membership
requirement over and above sharing the feature F and the interest [
(required by the definition). Otherwise, he won't be regarded by the
members as "one of us," and won't join in the group activity (or won't
be allowed to). A membership requirement is a convention in G of the
Appendix 282
form: If a person with F does A, then he becomes a member of G.
Sometimes A is going through a ceremony or receiving some kind of
initiation.
In the case of a social group being formed, its existence and member-
ship in it are a matter of mutual acknowledgment. This is true for a
primary group, where there is intimate face-to-face contact between
members throughout its existence. Members all know one another on
an individual basis, and it is the we feeling itself more than anything
else that determines membership. In this respect such groups as con-
sanguine families and athletic teams are borderline cases between pri-
mary and secondary groups.
Our definition of social groups should be consistent with the fact that
groups can persist even though the membership changes (this is true
mainly of secondary groups). The definition does not indicate how at
different times different sets of people could constitute the same group.
It seems that the main condition of identity over time is continuity of
membership. At the very least, this means that at no time after its
inception is it composed of all new members. In practice, groups have
names, symbols, meeting places, traditions, and the like to provide
continuity over time. In theory, however, there is no sure-fire way to
settle all questions of identity. It is easy to imagine situations where it is
impossible to decide whether the same group still exists or, after a
group divides into two, which (if either) is the original group and which
the splinter group. . sense of DF 11 , whereas influence relations are governed by informal
Organizations
A formal organization is much more structured than DF 16 requires a
social group to be. Its members are clearly differentiated by position, to
which specific duties and responsibilities are attached. Collective rules
(DFlO) with the status of regulations (DF 11 ) prescribe interconnecting
normative roles (DF 1S) that organize the activity of people in the same
and in different positions. Unlike in a social group, in an organi?:ation
the members need not have anything in common (other than being
members) or any common interest, and they need not share any we
feeling. Rather than identifying with the organization, they may be
motivated only extrinsically, by rewards or threat of punishment. 8 An
organization may have features of a social group, just as a social group
may have those of an organization (or contain an administrative organi-
Mutual Belief and Social Concepts 283
zation within it). DF 16 and the following definition of organizations are
not mutually exclusive:
DF17: G is an organization ex: there exist in G:
i. a differentiated set of positions defined by conventions, and rules for
filling them,
ii. collective rules determining normative roles for each position (duties
and responsibilities),
iii. collective rules for how and when to follow (ii) (procedures),
iv. lines of authority between positions and channels of communication
for exercising authority, and
v. conventions for recognizing (i) -(iv).
The first four clauses specify the distinct elements of structure that
constitute a formal organization, the rule-governed patterns of action
and interaction of people in differentiated positions with specified
duties and responsibilities (and procedures for carrying them out).
Strictly speaking, clause (iv), referring to lines of authority and chan-
nels of communication, is redundant, in that the rules specified in the
previous clauses must, in any real organization, cover authority and
communication. In order to differentiate official from unofficial pat-
terns of interaction, organization theorists often distinguish authority
from influence. In our terminology the official rules specifying author-
ity relations, indeed the official rules generally, are regulations in the
social norms and further determined by particular personal relation-
ships and norms of groups within the organization.
Clause (v) refers to what Hart (1961, 92ff) calls rules of recognition,
but since they determine what count as rules, we will call them con-
ventions. In an organization such conventions require not merely that
rules be mutually recognized but that they meet special conditions,
such as being enacted and codified. In effect, clause (v) stipulates a
mutually recognized means for determining what the rules are.
CONCLUSION
The analysis of basic social concepts attempted here is meant to eluci-
date their meanings and spell out their connections. The central con-
cept, mutual belief, refines the notion of consensus, which sociologists
have recognized to be central not only to the description of a great
Appendix 284
many social phenomena but also to the analysis of key social concepts.
By systematically defining these concepts in terms of mutual belief, We
make explicit their intersubjective character. The social reality of the
regularities and collectivities thus defined is constituted by people's
mutually dependent actions, beliefs, and attitudes. Thus, we need not
succumb to the reification advised by Durkheim of "considering social
facts as things." Nevertheless, we are in a position to understand why
social facts can strike people as things-their being socially real is
sufficient for that.
Apart from whatever theoretical understanding they provide, Our
definitions of basic social concepts have definite empirical possibilities.
They are formulated in terms of people's behavior and beliefs, both of
which are subject to empirical investigation. Moreover, the definitions
do not state if-and-only-if relationships, as philosophical analyses
generally do, but to-the-degree-to-which (cx:) relationships. Properly
quantified, these can become determinate functional relationships char-
acteristic of scientific propositions. The existence of norms, roles,
groups, and the rest is a matter of degree, as our formulations reflect,
and empirical investigation can determine the degree to which particu-
lar norms, roles, and so on, do exist and how that degree varies over
time, as they gradually come into and go out of existence, their scopes
broadening and narrowing.
Definitions are not theories and do not explain the phenomena whose
concepts they formulate. However, we think the concept of mutual
belief and the more general concept of the social distribution of belief,
including the cases of pluralistic ignorance and false consensus, have
great explanatory potential. Phenomena come to mind like socializa-
tion, person perception, public opinion, social cohesion, and mass ac-
tion, to name but a few topics of sociology and social psychology.
Notes
Notes to Chapter 1
1. In due course we will specify the ranges of the variables so-and-so, such-
and-such, and a certain way. Except for the term utterance act, our terminology
is Austin's (1962). Utterance acts for us are what Austin calls phatic acts, which
necessarily involve the performance of what he called phonetic acts, a notion
unnecessary for our purposes. Utterance acts involve producing certain sounds
belonging to (and as belonging to) a certain language, and are reported by direct
quotation. Austin characterizes the rhetic act as the use of a sentence" or its
constituents with a certain more or less definite 'sense' and a more or less
definite 'reference'" (1962, 93), and is reported by the familiar device of indi-
rect quotation. Although Austin speaks of the locutionary act as comprising the
phonetic, the phatic, and the rhetic acts, generally what he says about locution-
ary acts applies to them qua rhetic acts. For us locutionary acts are rhetic acts
in Austin's sense. See section 2.1.
2. A comprehensive treatment of this second issue would require solving some
very hard problems in the (general) theory of action-in particular, problems
of identity, individuation, and the part-whole relation of acts. We do not
propose (nor presuppose) a general theory, and the reader is invited to try to
subsume our discussion of speech action under such a theory, e.g. Goldman
(1970) or Thomson (1977).
3. Besides, as Austin noted (lecture X), there are uses of "by" that mark off
illocutionary, locutionary, and miscellaneous other redescriptions of an (utter-
ance) act.
4. A and B mutually believe that p if and only if each believes (1) that p, (2) that
the other believes that p, and (3) that the other believes that the first believes
that p. The idea (as "mutual recognition" and as "mutual expectation") origi-
nated with Schelling (1960, ch. 3), who also speaks of a matter of fact as being
"obviously obvious." For more discussion see the appendix.
5. Although there are some questions about how much leeway one has in
reporting what has been uttered (can we report "I vant a banana" as "I want a
banana"?), differences between utterances of the same sentence do not seem to
mqtter semantically. However, there is an utterance-act difference between
Notes to pp. 8-23 286
whispering something, speaking it, and shouting it, a difference between
mumbling something and articulating it clearly, and a difference between ut-
tering something slowly, normally, and fast; each of these utterance-act differ-
ences can make an illocutionary difference. A whispered "Leave" might be
taken as a plea, a spoken one as a request, and a shouted one as an order or
even threat. See Miller and Johnson-Laird (1976, sect. 7.4.1, and references
therein) for more discussion.
6. This point does not apply to certain special cases. If 8 and H are learning a
language together, it might well be that 8 utters e knowing H understands e just
because 8 has seen H use or understand e previously. Or if technical terminol-
ogy without wide currency is used, then special beliefs to the effect that His
acquainted with that vocabulary are required on the part of 8.
7. Even if, as a matter of empirical fact, people have to use or witness the use of
a language in order to learn it, our conceptual claim is not affected. Knowledge
of a language is one thing, its being mutual in a community is another.
8. For example, in saying that indirect quotation reports the proposition ex-
pressed by the utterance of the sentence, we do not mean to hide the philo-
sophically notorious problems of referential opacity and of the identity of
propositions.
9. Not all instances of the blank are related in quite the same way to the
interrogative uttered. "What time is it?" involves subject-predicate inversion,
whereas "Who discovered the calculus?" simply becomes: 8 says that H is to
tell 8 who discovered the calculus.
10. Not all illocutionary acts (greetings, for example) have propositional con-
tent, but we ignore this complication here.
11. Explicit performative utterances are no exception. As argued in chapter 10,
literally they are statements and only indirectly do they have the force of the
sort named by the performative verb. For example, a typical utterance of "I
order you to leave" is literally a statement and only indirectly an order.
12. Grice (1968, 60-61) seems to recognize this important difference when he
argues that with meaningful items, the reflexive intention can be replaced by a
simple intention. See also Schiffer's discussion of this point (1972, 133ff).
13. The reason for this choice of examples is the strategy of analyzing linguistic
meaning in terms of speaker meaning. See Grice (1968), Schiffer (1972, ch. 6),
and Bennett (1973, 1976).
Notes to Chapter 2
1. Depending on how 8 pronounces e, there can be problems in even reaching
Ll. We have made allowances (chapter 1, note 5) for utterances involving
mispronunciation, foreign accents, and faltering speech.
2. Pace Humpty Dumpty, there is no limit to what 8 can mean bye. However,
ordinarily if 8 means ... bye, 8 believes that e means ... in L (or at least in some
dialect of L shared by H).
Notes to pp. 25-35 287
3. This example is in the singular, but what follows applies, mutatis mutandis,
also to plural cases.
4. The way H represents to himself the referent (as "the such-and-such") may
in some cases be identical to the description 8 uses to refer to it (as the
"so-and-so") .
5. As Donnellan has observed (1966), there is another way in which definite
descriptions can be used to refer even when they don't apply to what they are
being used to refer to. 8 might utter "The diamond necklace looks magnificent
on her" and yet be referring to a cenain glass necklace. Even if 8 himself does
not believe the necklace to be made of diamonds, he may believe that H
believes this or that H believes that he (8) believes this, and so on. Whatever
the case, 8 succeeds in referring to the glass necklace if 8 and H mutually
believe of the glass necklace, however each thinks of it, that it is what 8 is
talking about. On our view, what Donnellan means by "referential" as opposed
to "attributive" uses of definite descriptions is not a special semantic feature of
definite descriptions but rather a special case of the general phenomenon of
successful communication despite discrepancies between what the speaker
says and what he intends to be thereby doing. Essentially the same point is
made by Castaneda (1977, note 7) and by Kripke (1977, especially 263-264),
who distinguishes "speaker's reference" from "semantic reference" and uses
this distinction to supplant Donnellan's.
6. Although the schema contains parameters for speaker, hearer, and expres-
sion (8, H, and e), it does not reflect variations in time. However, L3 could be
interpreted as having a time parameter as well:
L3. 8 is saying at t (to H) that *( ...p ... ).
By providing parameters for time, the schema is able to represent some uses of
the tenses and time reference: In uttering "I am tired now," 8 is saying at t (to
H) that 8 is, at t, tired. In uttering "You left," 8 is saying at t (to H) that H left,
at some time earlier than t. In uttering "(You will) leave!" 8 is saying at t, that
H is to leave at some time later than t. However, for the purposes of the schema
this time parameter may be omitted inasmuch as, at least in ordinary oral
communication situations, the time at which the hearer represents to himself
what the speaker is saying is approximately the time at which the speaker is
saying it. For other situations-hearing a tape or reading a memo-the time
parameter could readily be included.
7. These two interpretations are not intended to mark all the existing theories of
sense, some of which are discussed in chapter 8. As for theories of proper
names, among current rivals are the theories of Kripke (1972), Loar (1976), and
Katz (1977a).
8. Alternatively, 8 is saying that H is to make it be the case that H closes the
door. It is sometimes suggested that what be made the case is: the door is
closed. Although that might work for this example, it is very awkward for
others. Consider: in uttering "Give me two nickels for a dime" 8 is saying that
H is to make it be the case that 8 has two nickels for a dime.
Notes to pp. 36-38 288
9. Some acknowledgments (e.g. apologies) can be performed with constatives.
For example, uttering "I am sorry for that" can be used both to state that one is
sorry and to apologize. Our discussion of acknowledgments in section 3.4 will
make clear the difference. There seems, incidentally, to be no particular type of
saying associated with literally performed acknowledgments.
10. The main types of intonation that would have to be considered are con-
trastive stress and rising intonation on declaratives. It is not clear that the latter
should be treated like interrogatives, since they are not used to perform
analogous locutionary acts. It is very odd to say: In uttering "John admires
Vilas?" S was asking H whether or not John admires Vilas (or S was saying
that H was to tell S whether or not John admires Vilas).
11. Searle (and Katz by implication) denies the existence of locutionary acts
altogether, at least as Austin characterizes them. Searle remarks, "Every sen-
tence has some illocutionary force potential, if only of a very broad kind, bUilt
into its meaning. For example, even the most primitive of the old-fashioned
grammatical categories of indicative, interrogative, and imperative sentences
already contain determinants of illocutionary force. For this reason there is no
specification of a locutionary act performed in the utterance of a complete
sentence which will not determine the specification of an illocutionary act"
(1968, 412).
Searle is correct that every sentence has some illocutionary force potential
built into its meaning, and it is true that Austin neglected this point. However,
since any sentence can be used nonliterally in all sorts of different ways, what
Searle means must be restricted to the sentence's literal illocutionary force
potential. The specification of the locutionary act performed in the utterance of
a sentence used nonliterally does not determine the specification, even ap-
proximately, of the illocutionary act. Nonliteral utterances show that meaning
does not determine the identity of force. More fundamentally, meaning does
not guarantee the existence of illocutionary force. For example, without any
illocutionary intention S might say that Jesus was a Jew, with the perlocution-
ary intention of offending H. Thus, since a locutionary act can be performed
without any illocutionary act being performed, the distinction between the two
types of acts must be preserved.
Katz's (1977c) theory depends on ignoring this distinction. He believes that a
portion of speech act information should be represented in the grammar of a
language, that is, in the semantic component. Although he recognizes the con-
tribution of context to illocutionary force and that this contribution must be
handled by pragmatics (which is "performance theory at the semantic level"
(p. 15)), he thinks that semantic competence, "what an ideal speaker would
know about a sentence when no information is available about its context, that
is, when it is used in the 'null context'" (p. 14), includes considerable illocu-
tionary force information. However, contrary to what Katz says, the null con-
text is not an idealization like a frictionless plane, but a fiction, for there is no
situation, even ideally, in which context contributes nothing to what is being
done. When Katz says that sentence "meaning is the information that deter-
mines use in the null context" (p. 21), he implies that use is determined even
Notes to pp. 38-58 289
when (by his definition of "null context") no information is available about its
context. But clearly, if no such information is available, there is no such deter-
mination. At most, what is determined is the literal illocutionary force, if any,
but not that there is any. Perhaps because he does not acknowledge this point,
Katz develops a theory of propositional types which goes far beyond what we
include at the level of sentence meaning.
Notes to Chapter 3
1. In addition to Austin's and Searle's, taxonomies have been proposed by
Vendler (1972), Fraser (1973), McCawley (1975), and Schiffer (1972). Vendler's
and McCawley's are variants of Austin's; Fraser's and Schiffer's are substan-
tially different. Katz (1977, 195-222) classifies a variety of types of what he
calls "performative propositions" (a semantic category we do not use) and
claims that his scheme can be adapted to illocutionary acts.
2. Of course, the same verb may name several types of illocutionary acts, e.g.
"suggest" and "insist."
3. The analysis of advisories has the superficial form of that of constatives-in
p~icular, confirmatives; the difference is that in advising, S is providing H
WIth a reason to act rather than a reason to believe. Thus, like directives in
general, literal advisories are imperative in form. Advisories performed using
declaratives like "You should" or "It would be a good idea to" are performed
indirectly.
4. In his "How to derive 'ought' from 'is' ," Searle (1964; 1969, ch. 8) claimed,
quite correctly, that promises are not only acts of undertaking an obligation but
acts that create obligations. His view is that it is in virtue of being a successful
illocutionary act that a promise creates an obligation. This is apparently a
consequence of his theory of illocutionary acts as governed by constitutive
rules, which theory we discuss in chapter 7. On our view, the most that a
promise creates qua illocutionary act is the mutual belief between Sand H that
S's utterance obligates S to do A. That it does so in fact is another matter, be it
a matter of moral or of institutional fact. Whatever it is, it is not an illocutionary
fact.
5. Regarding this intention, the speaker's obvious insincerity results from the
obvious futility of trying to get the hearer to change his mind.
6. It might be objected that our broad interpretation of what it is to provide a
reason by one's utterance lets too much into the concepts of expressing an
attitude and of performing an illocutionary act, in particular, that it blurs,
indeed obliterates, the distinction between nonliteral and indirect illocutionary
acts by making seemingly direct, nonliteral acts into indirect acts. For example,
if S says to H, "I don't have any wheels," thereby telling H that he doesn't
have a car (available), clearly he is not telling H that he has no wheels, although
he would be if he were speaking literally. Yet, the objection continues, our
conception of expressing an attitude seems to imply that S is expressing the
belief that he doesn't have any wheels and thereby telling this to H. For his
Notes to p. 58 290
utterance surely provides a reason, rather than no reason, for H to think this
albeit a reason that is immediately overridden by the nonliteral reading of th~
utterance. If providing a reason is compatible with, indeed suggested by, obvi-
ous insincerity, then it would seem that S is being obviously insincere with
respect to the literal reading. Indeed, since he relies on this fact to get his
nonliteral intention recognized (compare an utterance of "I need to get my
wheels aligned"), there would be no basis for saying that S's utterance is
nonliteral rather than indirect.
This elaborate objection is based on a misunderstanding of just what our
conception of expressing an attitude is. In expressing an attitude, the speaker
R-intends his utterance to be taken as reason to believe he has the attitude. In
cases of obvious insincerity, even though S knows that his utterance will not be
taken as sufficient reason to believe he has the attitude, and knows that H
knows that S knows this (because there is a mutually believed reason to the
contrary), still S knows that his attitude toward the propositional content of his
utterance is at issue. Therefore, S can reasonably R-intend his utterance to be
taken as relevant to this issue. Even though he knows that it will not be taken as
(intended to be) sufficient reason to believe that he has the attitude in question,
still he can reasonably R-intend it to be taken as a reason, a reason which,
under the circumstances, is overridden by other matters of mutual contextual
belief. However, in the case of a clearly nonliteral utterance, when S says to H
(under suitable circumstances) that he doesn't have any wheels, his having or
not having wheels (literally) is not at issue and indeed is R-intended to be
recognized to be not at issue. H in turn infers (in accordance with L5 ' of the
SAS as elaborated in chapter 4) that S could not be telling him that he has no
wheels. Accordingly, H seeks a nonliteral reading.
Thus the case of direct nonliteral utterances has not been ruled out by our
conception of expressing an attitude. For it is precisely in this case that His
R-intended not to take S's utterance as reason to believe that S has the attitude
he would be expressing if he were speaking literally.
When S is speaking indirectly, he is expressing certain attitudes over and
above those expressed literally. Moreover, the hearer is to identify the atti-
tudes expressed indirectly in terms of those expressed literally. In other words,
he recognizes that S is F-ing that P by recognizing that S is F*-ing that p but
not merely F*-ing that p. In the case of direct nonliteral illocutionary acts,
however, H is intended to rule out S's F*-ing thatp and to recognize that Sis
F- ing that P instead. And H is to do this on the basis of what S is saying, not on
the basis of S's would-be literal act, F*-ing that p. When a person is speaking
nonliterally, he is using his words, some of them anyway, as if they meant
something different from any of their meanings in L, and he intends H to
recognize just how. In order to express an attitude other than one that would be
expressed literally, he intends his words to be taken differently from the way
they could be literally taken, in any of their senses, so that in effect he doesn't
mean what he says. But what he means instead must be identifiable, namely, in
terms of taking certain words differently in a way that is plausible under the
circumstances. This does not happen with indirect acts, because an indirect act
Notes to pp. 58-64 291
must be identified in terms of the literal act being performed; therefore, the
words of the utterance are to be taken literally.
If one still has misgivings about our way of handling the case of obviously
obvious insincerity, we offer an alternative formulation of what it is to express
an attitude, which we believe to be equivalent to, but more perspicuous than,
our original one. Instead of saying that expressing an attitude is R-intending H
to take one's utterance as reason to believe that one has that attitude, we can
say that it is R-intending H to take one's utterance as sufficient reason, unless
there is mutually believed reason to the contrary, to believe that one has that
attitude. In this way we can sidestep the issue of whether an overridden reason
is still a reason, while at the same time making it clear that communicating
attitudes is one thing and having attitudes is another.
The case of acknowledgments may seem troublesome even for our revised
formulation of the notion of expressing an attitude. This is because our defini-
tions already allow for perfunctory acknowledgments, which are obviously
insincere with respect to the feeling expressed. In these instances S intends his
utterance to satisfy, and intends H to take it to satisfy, the social expectation
that the appropriate feeling be expressed. Could an acknowledgment per-
formed with this intention be obviously insincere? Clearly it could be: for
example, S might be apologizing for something for which a perfunctory apol-
ogy is not enough. If this is mutually believed, then S cannot reasonably intend
his utterance to satisfy the social expectation that one express regret, nor could
he reasonably intend H to take it as satisfying that expectation. S would have
succeeded in apologizing, but H would surely not accept the apology.
7. Of course, he could be speaking nonliterally and be obviously insincere. In
the drinking example, instead of saying that he hadn't been drinking, S could
have said that he hadn't touched a drop; it would be mutually obvious that S
was speaking nonliterally, claiming, however insincerely, that he hadn't been
drinking.
Notes to Chapter 4
1. Though Grice (1975) was not the first to note all of these maxims of conver-
sation (see Harnish 1976b, 341), he was the first to bring them together as part
of a coherent and compelling theory of communication.
2. Recently R. Kempson (1975, 160-161) has attempted to formulate this
maxim more precisely, but the result is implausibly strong.
3. This accounts for Searle's generalization lover speech acts (1969, 65) with-
out appealing to constitutive rules.
4. In the case of questions, this amounts to not asking a question like "Have
you stopped beating your wife?" unless you have reason to believe what is
assumed.
5. Grice (1975) does not elaborate on the maxim "Be polite." Politeness is a
very tricky concept; part of the reason -may be that it acts in many ways like an
excluder (see R. Hall, 1959).
Notes to pp. 66- 88 292
6. Thus, it could be argued that understatements involve literal acts.
7. We have left open a number of issues surrounding RI-R3. For instance,
what exactly is the opposite of a term or expression, as mentioned in Rl? One
natural answer is that the opposite is the antonym. However, the notion of
antonymy is not much clearer than the notion of oppositeness and is often
characterized in terms of it. As for R2, various figurative relations fall under the
umbrella term metaphor. What counts as a metaphorical or figurative connec-
tion? These vary in perhaps unpredictable ways and we have mentioned only
two of the types that have received labels:
(R2) a. Synecdoche: a salient, distinctive whole/part if a part/whole is
mentioned.
b. Metonomy: a salient, distinctive associated object, property, or
relation.
8. Note that in this case of sarcasm negation goes to the embedded sentence,
and is thus not simple external negation.
Notes to Chapter 5
1. Because the LP and the CP are applicable to normal communicative situa-
tions, the mere fact of utterance warrants H to think that S has some communi-
cative intention identifiable by what is uttered. Only where the CP (or the LP)
is inoperative is it necessary for H to infer from the nature of S' s utterance the
existence of some communicative intention. In effect, then, we are saying that
linguistic communication, at least as normally accomplished, requires the LP
and the CPo Nonlinguistic communication does not and thus requires a more
complex inference by H. On some occasions (see note 4) nonlinguistic commu-
nication can be performed using language.
2. It is not clear what such a claim would amount to anyway. Rational re-
constructions (in science and philosophy) are usually axiomatizations of less
explicitly formulated working theories. Conditions of adequacy on such an
enterprise include preservation of theorems, minimal redundancy in primitive
terms, etc. None of these characteristics is appropriate to the present case,
unless nothing more is meant by "rational reconstruction" than 'idealization.'
See Suppes (1957, ch. 12) and references therein for more discussion.
3. See Fodor (1975, ch. 3) for a discussion of these kinds of cases from the point
of view of cognitive psychology.
4. Suppose during the rehearsal of a play, wherein the CP is suspended, S,
following the script, says to H "Tonight is the night, my love." H's husband
happens to be the director, and it happens that he, as mutually believed by S
and H, will be spending the evening with the producer. Moreover, there. is a
certain special twinkle in the eye of S that only H can observe. In this context,
despite the suspension of the CP, S could reasonably have uttered the line with
the communicative intention that H take it as an invitation to see him that
night. H relies not on the CP but on the twinkling in S's eye to infer that S has
some communicative intention.
Notes to pp. 89-114 293
5. Many of the published counterexamples to the various versions of Grice's
account of speaker meaning rely on there being no CP operative (see Strawson
1964; Grice 1969; Schiffer 1972). What is in question in those cases is not the
identity but the existence of S's communicative intention. These examples
generally involve nonlinguistic communication, where there is nothing about
the "utterance" analogous to being a sentence of English that activates the CP.
For this reason, we suggest, discrepancies between the intended inference and
the inference made produce counterexamples-cases that don't seem to be
instances of communication. Finally, there are linguistic counterexamples;
even if they are cases of communication, they are not linguistic communica-
tion, since S does not intend H to rely on the meaning of the words he utters. In
Searle's oft-discussed example (1969, 44f) of "Kennst du das Land wo die
Zitronen bliihen?" used to communicate that S is a German soldier, S intends
H merely to rely on the fact that the sentence is in German.
6. In general S also has a perlocutionary intention, recognition of which would
further explain his utterance. However, perlocutionary intentions need not be
recognized to be fulfilled; therefore H need not be able to explain S' s utterance
at this level.
7. It is not, to use a catch-phrase in contemporary philosophy, an inference to
the best explanation (see Harman 1973). Rather, H takes the fact that he has
found a plausible explanation of the speaker's utterance as good reason to
believe this explanation to be the right one.
8. The inference is usually made routinely, but when the speaker is particularly
obscure or subtle, explicit inference may be necessary before the hearer is
confident (if he ever is) that he has things right.
9. For example, S might ask H "What are you doing tonight?" intending it to be
taken as an invitation only if the answer is "Nothing" or something to that
effect.
Notes to Chapter 6
1. See section 7.1. Rawls' (1955) practice conception of rules and Searle's (1969)
constitutive rules are similar to our conventions. Unfortunately, their accounts
do not clearly distinguish rules as requirements from the conventions defining
the actions required. Another possible confusion is to think that every act that
is part of a ritual, ceremony, or other formalized procedure is conventional. Of
course it is in one sense, but not necessarily in the relevant sense. Such an act is
conventional in the relevant sense only if the description it falls under makes
reference to institutional facts such as the position of the act in the procedure.
2. These are essentially the same as Searle's (1975b) declarations and repre-
sentative declarations, although for him all illocutionary acts are conventional
in the sense of being governed by constitutive rules. His labels are slightly
misleading, since representative declarations are not declarations (on his ac-
count).
3. Strictly speaking, we should say that an institutional fact is anything ex-
pressed by a proposition that is true in virtue of being mutually believed in
Notes to pp. 114-126 294
some collectivity, or anything expressed by a proposition that follows from a
proposition expressing one or more institutional facts (possibly together with
other true propositions).
4. Even if some self-made autocrat makes the basic decisions that determine
the major institutional facts in his society, these institutional facts, including
the fact that what he says goes, are still matters of mutual belief. It is not a
political but a conceptual point that the existence of these facts (their describa-
bility in institutional terms) depends on mutual belief, even if one person has
the power to cause these facts to obtain.
5. We don't mean to imply that every fact about an institution is an institutional
fact. To be sure, every fact describable only in terms that are applicable in
virtue of mutual belief (whether about acts, persons, procedures, policies,
rules, or objects) is an institutional fact. However, what we have in mind as
noninstitutional facts are those that might describe social cohesion, social mo-
bility, stability, power distribution and various other matters of concern to
sociologists. In general, sociological facts about institutional facts are not in-
stitutional facts.
Notes to Chapter 7
1. Of course, here we are not using "express" in our technical sense of'R-
intend the hearer to infer.' In our imaginary situation, the speaker has a simple
intention for the hearer to infer his attitude on the basis of his utterance's falling
under a communicative convention of the sort we are imagining.
2. Still, it might seem that the very utterance of appropriate words in the right
circumstances automatically counts as an acknowledgment, regardless of S's
(R-)intention. Greeting is especially troublesome. It would seem that to say
"Hi" or "Hello" upon encountering someone for the first time on a given
occasion could not but count as a greeting and therefore does not require an
R-intention or recognition thereof. Indeed, no intention at all seems required.
But suppose that H believes S is not aware of him and so cannot be acknowl-
edging his presence. Since uptake has not been achieved, has S succeeded in
greeting H? Uptake involves attributing to S a communicative (R-)intention.
And if S realizes that H believes he (S) is unaware of H, S cannot expect his
utterance to be taken as a greeting. To be a greeting, an utterance of "Hi" or
I'Hello" must at least be addressed to someone and be intended to be taken as
so addressed. But that is hardly enough. Part of what makes such utterances
seem automatically to be greetings, irrespective of intention, is that it is difficult
to imagine nonliteral uses of such terms. But it is not impossible. Sand H might
not be on speaking terms. For days they pass each other without exchanging a
word. Finally, tired of sidelong glances, S says "Hi" to H. Relying on the
mutual contextual belief, which is likely to have developed by this time, that
their dispute was too petty to be permanent, S R-intends his utterance to be
taken as a proposal to resume normal relations. He relies also, of course, on the
standard use of "Hi" to greet friends but not necessarily enemies. We con-
Notes to pp. 126-140 295
elude, then, that there is more to greeting than saying "Hi." Even a per:"
functory greeting is R-intended to be taken as such.
3. Strictly speaking, for Searle only some constitutive rules take this form,
which he contrasts with the characteristic form of regulative rules like" 'Do X'
or 'If Y do X'" (1969, 34). Mysteriously, he says in the very next sentence,
"Within systems of constitutive rules, some will have this form," but he never
explains why rules (within a system of constitutive rules) that take the form of
regulative rules are not themselves regulative rules, albeit regulative rules that
govern the behavior defined by constitutive rules (see section 7.1). Accord-
ingly, our discussion focuses on constitutive rules that take the counts-as form.
4. Searle notes this feature (1969,61, note 1), but does not seem to consider it as
severe and artificial a limitation as we think it is.
5. This seems to be a consequence of the if-clause in R(g). Supposing that
conditions 1-8 do obtain, R(g) requires that Sutter e if S is to speak sincerely
and correctly. But suppose that (in the dialect) some other sentence e' means
the same thing as e and (so) has the same semantic rules as e. Then conditions
1-8 will be sufficient for the correct and sincere utterance of e' also. Since 1-8
define what it is to promise (in uttering something so used), it seems to follow
that if one is to promise in this dialect one must utter e and utter e' . Also, the
only-if clause leaves the problem of ambiguous sentences unresolved. If a
sentence is ambiguous, then on Searle's account the utterance of e could have
incompatible necessary conditions if p were true and q were false:
i. Utter e only if p.
ii. Utter e only if q.
Under such circumstances one could not utter e without violating a rule of the
language.
6. For instance, Grice (1969) has this as a consequence of a principle governing
intentions to the effect that one must suppose that there is at least some chance
that one's intention will be fulfilled in order to have that intention.
7. He does analyze a notion of convention relevant to coordination problems
and points out that this is not the sense in which utterance types are conven-
tional. By the way, his formulation of this other notion of convention comes
elose to our notion of a practice (see appendix).
Notes to Chapter 8
1. See Kempson (1975, ch. 8) for some excesses of this sort.
2. It might use a simple, unambiguous syntax and subscript all ambiguous
lexical items.
3. Given this connection between decomposition and prediction of semantic
properties and relations, it was very prudent of Davidson (1967) as a non-
decompositionalist to deny the necessity of accounting for such properties/
relations as ambiguity, synonymy, entailment, etc. His subsequent (1970) re-
versal on some of these matters jeopardizes his nondecompositional position.
Notes to pp. 143-154 296
4. The merits of inference over decomposition from the point of view of psy~
chology of language will be discussed further in chapter 11.
5. Fodor, Bever, and Garrett (1975, 180) give the following as a sample of What
they call a "rule of inference": "A formula of the form P and Q entails a
formula of the form P and a formula of the form Q." But this is not a rule of
inference, it is a statement about a semantic relation.
6. This is not necessarily to say that it has some other form or forms, at least on
the view that only sentences that are (or express something that is) true or false
have a logical form.
7. Moreover, what it is for an expression to have a meaning is not specific to a
language, though grammars are.
8. "Mean" can also be used for 'intend to refer to' as in "He meant Connors
when he said he met a famous tennis player."
9. Grice himself seems to have conflated these last two uses in his original
(1957) article, but distinguishing them (Grice 1969) renders his analysis immune
to criticisms like Ziff's (1967).
10 ..When not speaking literally or not seriously (see (1) and (2)), for S to mean
(5) that p may be for him to imply that p.
11. Our points about simple sincerity and about intended hearer response can
be strengthened by being tied together. Whereas any form of subtle deception
involves the speaker having an intention he intends the hearer not to recognize
or not having an intention he intends the hearer to think he has, simple decep~
tion is straightforward insincerity-the speaker does not believe what he
seemingly wants the hearer to believe, or he does not want the hearer to do
what he seemingly wants the hearer to do. Now it will not do to say that the
notion of speaker meaning does not preclude insincerity, because insincerity is
not meaning what one says in the wrong sense (1) of the word "mean." At least
it won't do if, as Schiffer argues, to mean something is to intend to cause, in the
way specified by Shiffer's analysis, H either to actively believe something or to
do something. And these are, according to Schiffer (1972, 80-87), the only
types of response that are tied to meaning something. Nothing in Schiffer's final
analysis (1972, 63) precludes insincerity, and yet according to that analysis S, in
meaning something, issues an utterance with the intention of realizing a certain
state of affairs E that, if realized, is sufficient for Sand H to mutually know that
E obtains and that it is conclusive evidence that S has the primary intention of
causing H to believe or to do something (in the way specified by the further
details of the analysis). But how can S reasonably expect his primary intention
to be fulfilled if H thinks S is insincere? To be sure, Schiffer's analysis requires
that H's response be for an intended reason, and that this be achieved, at least
partly, by H's belief that S's utterance be related in a certain way to the
intended response (these requirements are given in clauses (1) and (2) of Schif-
fer's analysis), so it might be thought there wouldn't be a reason if H thought S
was insincere. In this case, Schiffer's analysis implies that S at least intends H
to have taken him to be sincere. However, by not being explicit about S's
Notes to pp. 154-155 297
sincerity, or at least about S's intention for H to take him to be sincere,
Schiffer's analysis obscures the fact that the only way S's utterance can cause,
in the required way, H to believe something or to do something (except for
cases like reminding) is for H to take S's utterance as evidence that S believes
that p or that S wants H to '1'. Otherwise, the intended connection that S
intends H to recognize between his utterance e and the intended response will
only reveal S's insincerity, not incline H to respond as intended. Schiffer's
analysis fails to make explicit why H actually responds as he does, when he
does as he is meant to do. H's reason for so responding is based on what he
thinks S thinks or wants, not merely on deciphering S' s primary intention about
what H is to believe or do. However, once we make explicit that H's response
(except for reminding cases) is mediated by his attributing a certain belief or
want to S, it is clear that S's intention that H attribute to him (S) a certain
attitude is distinct from his intention that H respond in a certain way. It then
becomes clear that once H identifies what S believes or wants, the primary
intended response is clearly perlocutionary. For once H identifies S's belief or
desire, as expressed by S's utterance, no recognition of any further R-intention
is required for H to respond appropriately to that belief or desire, which is the
primary response required by Schiffer's analysis. Instead, H responds as in-
tended, reasoning as follows:
Declarative cases:
1. S believes that p (by recognition of R-intention).
2. S wants me to believe that p.
3. Therefore, I will believe that p (unless there is strong reason to the contrary,
i.e., unless I am not convinced).
Imperative cases:
1. S wants me to 'I' (by recognition of R-intention).
2. Therefore I will 'I' (failing a strong reason to the contrary).
Recognition of S's R-intention figures only in step 1. H's primary response is
not a further matter of recognizing S's R-intention, for there isn't any further
R-intention to recognize.
12. Besides offering various informal conceptions of presupposition, numerous
authors fail to apply their pet conceptions consistently to their own examples
and even shift conceptions within (and between) articles without notifying the
reader. As a result there is a variety of ways to categorize these observations:
lexical vs. sentential, sentential vs. speaker, semantic vs. pragmatic, logical vs.
psychological, etc., depending either on the examples, the conceptions of pre-
supposition espoused, or the conception of presupposition that actually applies
to the examples.
13. Frege did not rigidly maintain a distinction between what a SMU of a
sentence presupposes (vs. what a sentence means-its sense) and what a
speaker presupposes (vs. what a speaker asserts). Frege contrasted presuppo-
sition with both sense (hence with entailment) and assertion. It was not neces-
sary for his purposes to keep these distinct since for Frege, what a speaker
asserts and the content of a declarative sentence are (indexicals aside) the same
Notes to pp. 155-162 298
thing: the thought that p. (See Harnish 1977, and in preparation, for more
discussion. )
14. Strawson (1950) also seems to have entertained the doctrine that no state-
ment is made if the presuppositions are not satisfied, but this theory would not
accord well with our definition of asserting in terms of expressing propositional
attitudes, nor with the plausible view that a statement can be what is asserted.
15. Frege actually says (1892, 69) that (13a) presupposes that the name "Kep-
ler" refers to something, which conflicts with other doctrines he holds and
conclusions he draws. (See Hamish, in preparation, for further discussion.)
16. Strawson (1952, 176) continues by saying that in general, "There are many
ordinary sentences beginning with such phrases as 'All . . . " 'All the . . .',
'No .. .', 'None of the .. .', 'Some .. .', 'Some of the .. .', 'At least one of
the ... ', which exhibit, in their standard employment, parallel characteristics
to those I have just described in the case of a representative' All ... ' sentence.
That is to say, the existence of members of the subject-class is to be regarded as
presupposed (in the special sense described) by statements made by the use of
these sentences; to be regarded as a necessary condition, not of the truth
simply, but of the truth or falsity, of such statements."
17. Not every negation forms a contradictory. Usually internal negation creates
a contrary, not a contradictory sentence. Whereas an external negation such as
"It is not the case that ... " always does. Thus external negations should be
used in tests calling for contradictories. The contradictory of s is not-s, true if
and only if s is false.
18. Keenan's example does not seem to fit his characterizations. It is not
obvious that a perfect stranger (adult) of equal social status could not say (in
French) to H "Tu es degoutant" after seeing H spit on the sidewalk, and be
taken to mean literally exactly what he said.
19. From S's standpoint sometimes the relation can be weaker; S believes H
believes that p and S believes that H believes that S believes that p.
20. The external negations give a similar, but weaker understanding; compare
(31b) and (32b) with "It is not the case that John realizes that his car has been
stolen" and "It is not the case that John has stopped playing tennis." Perhaps
when the negative element is placed next to the verb "realize" (internal nega-
tion) it tends to attach more firmly to the belief condition because it is that
condition that adds new information to the complement of the verb- the com-
plement could itself be negated if that were all one wanted to communicate.
When the negative element is distant from the verb, the conditions are more on
a par and thus the negation can go to the conjunction, though the belief condi-
tion is still favored.
21. It seems, incidentally, that the theory of semantic presupposition cannot
handle these cases without postulating ad hoc ambiguities.
22. See section 7.1 for the definition of a rule.
Notes to pp. 163-166 299
23. It might be thought that there is a way of taking (46) so as to be con-
sistent-assume that x is completing someone else's -ing. But (43b) cannot
really be used to communicate that.
24. See Hamish (1976b, 373-389) for some discussion of additional cases from
this perspective, and D. Wilson (1975, ch. 6) for additional data.
25. This diagram is meant only to be suggestive; no readily apparent interpreta-
tion of the domination lines is coherent. It is an interesting and important
quest~on how these notions are related. We will not discuss Grice's theory of
meanmg any further, but it is important to know how utterer's meaning and
utteranc~-type meaning are related to saying and implying. Especially vexing is
the relatIOn between utterer's meaning and implicature. We would argue that
what one implicates (as well as what one says) is always a species of what was
meant, differences being traceable to different ways in which various intentions
are to function and different reasons the audience is intended to have. It is
important for this strong view that one not be able to imply that p and not mean
that p, from which it follows that one could not imply unintentionally.
26. The reason for the disjunction is that Grice's terminology has shifted from
the first of each pair to the second. The shift to implicature frees him from the
restrictive logical use of implies, and the shift to said is a generalization from
state. One can report what someone said without reporting what was stated for
instance, when one reports that he said that you are to leave the room. '
27. In an early paper (1961) Grice discusses four cases "in which ... some-
thing might be said to be implied as distinct from being stated." Although his
main concern is to decide on the vehicle of implication in each case, Grice does
suggest a pair of useful diagnostic tests (p. 446):
Nondetachability (of the implication from what is asserted): "Any way of as-
serting what is asserted ... involves the implication in question."
Noncancelability (of the implication without canceling the assertion): "One
cannot take ... [another] form of words for which both what is asserted and
what is implied is the same as [the first] ... and then add a further clause
withholding commitment from what would otherwise be implied, with the idea
of annulling the implication without annulling the assertion."
Suppose S has said that p, and we wonder whether S has thereby said or
implied that q. If S can consistently claim p, but not q, then S did not say that
q, so but could only have implicated it.
28. Some have been misled by Grice's remarks. For instance, Elinor Keenan's
interesting paper (1976) is somewhat marred by the supposition that her data
from Madagascar are counterexamples to Grice's theory of conversation. She
assumes that for maxims to be universal, they must be categorically observed.
But this need not be so. Grice's theory of implicature requires just that the
speaker and hearer(s) be observing the cooperative maxims. This assumption is
necessary for Grice's theory of implicature to work, but it is also sufficient.
Grice nowhere says, nor would want to say, that all conversations are governed
Notes to pp. 166-170 300
by the cooperative maxims. There are too many garden-variety counter-
examples: social talk between enemies, diplomatic encounters, police interro-
gations of reluctant suspects, most political speeches, and many presidential
news conferences. These are just some of the cases in which the maxims
of cooperation are not in effect and are known not to be in effect by the
participants, notwithstanding pretenses to the contrary.
Since Grice's theory is basically conditional (if any conversation is governed
by the cooperative maxims, then implicatures can be explained in a certain
way), falsifying the theory takes more than showing the antecedent sometimes
to be false. What must be done to falsify the theory, and what Keenan does not
report doing, is to find examples of cooperative exchanges where the maxims
are violated in the requisite way and none of the predicted implicatures are
present. Keenan does ask the important question: why don't the Malagasy
cooperate with information? Apparently the Malagasy are so closely knit that
possessing information another lacks gives one status over the other; reluc-
tance to cooperate on information could be a natural consequence of the gen-
eral reluctance to relinquish advantages in status.
29. Fragmentary accounts of various maxims can be found in the literature. On
quantity see Strawson (1964b), on quality see Ross (1930, 21) and Urmson
(1952, 224-230). This fact does not diminish Grice's achievement of widening
these observations and integrating them into a theory.
30. The maxim of relation (our presumption of relevance) turns out to be so
central and important in conversational implicature that it is not clear that it
belongs on equal footing with the others. We suspect that maxims are (at least
partially) ordered with respect to weight, etc., and that relevance is at the top,
controlling most of the others.
31. The general question arises of how to discover new maxims and how to
justify the claim that some maxim governs discourse. On our view of course,
conversational presumptions (maxims) have the status of contextual mutual
beliefs. As such they contribute to the explanation not only of implications but
also of other conversational phenomena. See chapter 4.
32. We want to emphasize that these are not mutually exclusive ways of in-
fringing a maxim. Clashes seem to be a special case of violation in which the
violation is forced, or seen as forced, by maxims that conflict under the cir-
cumstances.
33. In this reasoning H makes use of at least the following information: (1) the
conventional meaning of the words, (2) the identity of the referents, (3) the
conversational maxims (or CooP), (4) the context of utterance, (5) background
knowledge, and (6) the mutual belief that Sand H share knowledge of (1)-(5).
These items correspond to items in the SAS. Items (1) and (2) correspond to L2
and L3 of the SAS; item (3) to the communicative presumption; items (4) an~
(5) to the mutual contextual beliefs (MCBs) cited in the SAS. Item (6) makes It
clear that the other items are matters of mutual belief.
34. Notice that the explanation cannot turn on a difference in the verbs, be-
cause the same verb in different sentences can have different implications (see
Harnish 1976b).
Notes to pp. 170-1S0 301
35. Clause (4) is not clearly true and seems to conflict with (2). In (4) Grice
seems to be invoking his old criterion of cancelability as the test for whether
what is said is the vehicle of the implication. But since (nonconventional)
implicature is not entailment and can always be canceled, it is not clear why
truth and falsity are relevant.
36. It was this narrow, strongest notion of conversational implicature that we
used in effect in section 4.3.
37. Grice (1975,51) seems to deny this by classifying two examples ofimplica-
ture as "examples in which no maxim is violated, or at least it is not clear that
any maxim is violated." However, one paragraph later he writes, "In both
examples, the speaker implicates that which he must be assumed to believe in
order to preserve the assumption that he is observing the maxim of relation." If
so, then the speaker has infringed the maxim of relation at the level of what is
said. We think that Grice's latter characterization is right, and that the exam-
ples differ from the flouting examples only in the degree of irrelevance.
3S. And more generally, what is it for e to mean something?
Notes to Chapter 9
1. See Sadock (1972). This locution suggests these senses are not composi-
tional, but Sadock himself proposes underlying structures for them similar to
those for imperative sentences literally used to make requests, and these are
surely compositional. One problem is that it is not clear whether, or how,
generative semantics can define compositionality-the meaning of a complex
expression being a function of the meaning of its meaningful constituents and
their grammatical relations.
2. Sadock modifies this eventually (1974, 114) to "I indirectly request you
to ... " But since indirection (on Sadock's account) consists of a disparity
between form and function it is unclear what this could amount to as a piece of
semantic representation.
3. Question mark indicates rising intonation on the tag.
4. Exclamation point indicates falling intonation on the tag: "The tag has ap-
proximately the same intonation as the imperative kiss me" (Sadock 1974, 133).
5. The sentence may be grammatical if "please" is read as short for "please tell
me." But in that case "please" does not modify the verb denoting the requested
action in the main clause-contrary to the generalization used.
6. But notice that in spite of (Sd) and (14b), we do have "When will someone
please wash the car?!"
7. But we do get" When are you going to buzz off?!" In general the more vulgar
forms ("fuck off") work better in this argument.
S. Following Green (1972), Sadock applies the term to sentences used for
issuing directives in general. We limit it to the three forms given earlier. (Here,
however, we will ignore the third form: "How(s) about ... ?")
Notes to pp. 181-195 302
9. Another use of such constructions as "Step on my other foot, why don't
you?" is sarcastic and not literal.
10. For some people it is progressively more odd to say ?"Yeah" and ??"Yes."
11. If this is correct, "Can we go now?" as a suggestion will have to be indirect
off of the question reading.
12. Searle (1975a, 67-68) misses this point. An ambiguity theorist will argue
that a second meaning is a necessary part of the best available explanation of
certain linguistic facts.
13. This seems to be the view of Searle (1975a, 64, 68, 76, 77) who says, "there
can be conventions of usage that are not meaning conventions" (76). Since
Searle has no analysis of conventions (other than as systems of constitutive
rules, which he cannot invoke on pain of collapsing conventional meaning and
conventional use), it is impossible to evaluate this proposal on his terms; con-
sequently, we evaluate it on ours.
14. Presumably the speaker is not force-punning, simultaneously using both
readings.
15. There are some internal difficulties. For instance, "?(the hearer can pass the
salt)" is not the same proposition as "the hearer is to pass the salt"; thus the
formulation would have to provide machinery for forming the propositional
content of the standardized illocution.
16. Think of ordering paradoxes, or the Bach-Peters paradox (see E. Bach
1970).
17. We are not here addressing the question of characterizing the various types
of sentences that instantiate T. This is a research problem for any account of
illocutionary standardization. Whether speakers have to know extra meanings
(as the ambiguity thesis requires), illocutionary conventions (as the conven-
tionality thesis requires), or merely illocutionary precedents, whatever this
knowledge is it must somehow include a specification of the types of sentences
it concerns.
18. By dropping (ii) we get the related notion of having a standardized use for
F-ing that p. An expression can have a standardized use without ever actually
being used that way-though it could be so used. For instance, it could very
well be the case that no one has ever actually used a long compound sentence
like "Could you change the tint on the TV, and turn down the boiler on the still,
and ... ?" Yet in appropriate' circumstances we would know what was being
done in the utterance of such a sentence. This said we can concentrate on the
stronger notion of being standardly used.
19. This does not even appear to be the case with standardized forms used
literally to express the speaker's desire. An indirect request made by uttering
"I'd like you to take me home now," is also, uncontroversially, a statement of
the speaker's preference. Clearly the ambiguity thesis could not be plausibly
suggested for cases like these.
20. See our discussion of this point in section 3.6.
Notes to pp. 196-201 303
21. In some ~ases 1- may be a ,?ecessary condition for doing what is indirectly
r~quested, A .. S mIght ask H, Can you reach the buzzer?" in order to request
hIm to push It.
22. "You canA" can also be used as a permissive, but then "can" means 'may'
rather than 'is able.'
23 ..These forms don'.t take the preverbal or postsentential "please" because the
actIOn wh~se .ne~essity or compulsoriness is literally questioned is the negation
of the actIo~ IndIrectly requested. "Must you A?" is an indirect request for H
not to (cont~nue to) A, and "Mustn't you A?" is an indirect request for H to A.
Therefore, I~ "please': is read as "if i~ pleases you (to do what I'm requesting
you t~ do), the actIon referred to IS not the one mentioned in the literal
ques~~on. The s~e .point applies to literal questions like "Can you reach the
salt? where the IndIrect request is for the hearer to pass the salt (*" Could you
reach the salt, please?"). '
24. In our discussion ofillocutionary modifiers and of syntactic liberties (chap-
ter.10) we generalize our claim about "please." We argue that the grammati-
cru.Ity of a sentence is not guaranteed either by its being acceptable or by its
beIng usable. A ~ramm~ cannot do everything-if certain linguistic phenom-
ena c~ be explaIned withou~ ~emanding contortions from the grammar, they
needn. t be regarded as pertaInIng to grammaticality. This does not mean that
anythmg one's ~~mar doesn't explain is not a matter of grammaticality-
there must be prInCIpled reasons for construing such a phenomenon as not one
?f gr~mar. In the case of illocutionary modifiers, we argue that the sentences
m.which they occ~r are not grammatical although they admit of readily deter-
mmabl.e conversatIonal paraphrases by sentences that are fully grammatical.
We pomt out that the higher performative approach could work for only some
of these sentences anyway, for that only in some cases do the conversational
paraphrases take the form of structures with higher performatives.
25 ..It se~~~ th~~ the occurrence of "please" is restricted to sentences whose
subject IS y?u .and whose verb phrase is in the simple present or simple
future. Thus, Instead of (54) and (55), the following examples wouldn't illustrate
the right point: *"Did you please pass the salt?" *"1 can please pass the salt."
26. A further problem for the ambiguity thesis is presented by interrogatives
used as requests for permission. Consider (a) "May I (please) go now?" With-
?ut "pl~ase" this example can be used literally as a question asking whether S
IS permItted to go. It needn't be used to request permission, for H might not be
the o.ne. empowered to give it. However, (a) can be used as a request for
pern~.1Ss.IOn as ~ell, and with "please" it could only be so used. No doubt an
ambIgUIty theons~ ~ould say that it derives from an underlying imperative that,
on t~e performatIvist story, would look something like "I request that you
p~rmit me to go no~<' ?~ that a~count "please" would be grammatically no
different from what It IS In ImperatIve and whimperative contexts. On our view
an u~~eranc~,of (a) is a literal question and an indirect request for permission,
and please must be handled paragrammatically.
Now consider (b) "May I (please) have a beer?" Its utterance (with or

Notes to pp. 201-206 304
without "please") could be either an indirect request for H to give S a beer or
an indirect request for H to permit S to get a beer. As a literal question it is
unambiguous. The problem for the ambiguity thesis is to account for both
indirect uses. Does the sentence have three meanings? It seems that these
examples further illustrate the difficulty of trying to incorporate the "please"
phenomena into grammar.
Finally, compare (b) with (c) "I'd like some beer, please." Without "please,"
used literally (c) would be a statement of S's preference, and with or without
"please" it can be either an indirect request for H to give S a beer or an indirect
request for H to give S permission to help himself. For the same reason we
denied that (b) has two meanings, we deny that (c) has three.
27. Notice that an indirect report of (52) could not include "please," just as, in
general, illocutionary modifiers cannot be included in indirect reports of what
the speaker said: *"He asked loe me whether I could please pass the salt."
Notes to Chapter 10
1. For instance, Sadock (1974, 12) says "No theory of grammar can completely
avoid treating illocutionary force because of the existence of explicit perform-
atives."
2. The first-person plural can be used performatively, as when a spokesman
speaks for a group; and the second-person passive can be used performatively,
as in "You are commanded ... " We follow Austin's general usage of "per-
formative" as meaning explicit rather than primary performatives, a distinction
he "introduced rather surreptitiously" by p. 69.
3. Such sentences need not be used performatively, as when one is speaking in
the historical present or describing one's habitual behavior. See Austin (1962,
64, 68).
4. Most English verbs normally take the progressive (continuous) present,
rather than the simple present. Zeno Vendler (1972, chs. 1, 2) notes that like
performatives, statements of mental acts and of mental states also take the
simple present. He draws some interesting parallels between them and per-
formatives. See also discussions by Katz (1977c) and Ginet and Ginet (1976).
5. We are sticking with Austin's provisional performative/constative distinc-
tion. His subsequent general theory of illocutionary acts collapsed that distinc-
tion, since constatives are performative in the sense that they involve doing
something. But they are also true or false ("I state that ... "). Still, he never
came to hold that all performatives can be true or false.
6. That this regress does not follow is argued in Hamish (1976a).
7. Another flaw is the vagueness of" something slightly less than full conven-
tional force."
8. Austin (1962,69-70) gives a similar variety of cases to show what "making
explicit" conveys, but he claims that a performative is not a statement only
because it could not be true or false. This passage is the closest he comes to an
argument.
Notes to pp. 207-221 305
9. Katz (1977, 175f) rejects this supposition (hence our argument) because he
thinks performativity must be explained semantically, not pragmatically. (See
also our note 11, ch. 2.)
10. Austin (1962, 57) uses a legalistic locution to describe the force of "hereby"
as serving to indicate that the utterance "is, as it is said, the instrument effect-
ing the act."
11. Lemmon (1962) includes performative sentences among those to which his
title, "Sentences Verifiable by their Use," applies. We applaud his recognition
by implication, of the constative character of performatives, but we would
characterize them, unlike most of Lemmon's other examples, as "true in virtue
of their use."
12. This claim does not apply to conventionalized performatives.
13. Fraser indicates that hedged performatives are to be accounted for differ-
ently from simple performatives when he points out that the speaker can cancel
the illocutionary force of hedged performatives but not of simple performa-
tives. Compare "I must ask you to leave, but I won't" with "I hereby ask you
to leave, but I'm not (asking you to leave)." Only in the second case, involving
a simple performative~ does the speaker contradict himself. On our view this is
accounted for by the fact that he is both asserting and denying that he is asking
the hearer to leave. However, in the first example, the speaker is asserting that
he must ask, not that he is asking, the hearer to leave, whereas he is denying
that he has any intention of asking the hearer to leave. Clearly there is no
contradiction in this case. Our account of hedged performatives will thus be
different from our account of simple performatives. For example, in "must"
cases the speaker's statement is not of what he's doing but only of what he must
do.
14. Fraser claims (1975, 196) that in hedged performatives the "use of 'must'
implies a sense of helplessness" with respect to what the speaker says he must
~o: he is doing it not because he wants to but because he has to. Fraser's point
IS that the speaker in using "must" is trying to relieve himself of some of the
o~us of responsibility for what he is doing. But compare the hedged request (7)
WIth the hedged admission "I must confess that I forgot your name." To cover
both sorts of cases we suggest describing the speaker as reluctant. Being help-
less because obligated is but one route to reluctance. One can be reluctant to
harm the hearer but also reluctant to harm (embarrass) oneself.
15. Insofar as the locutions are of standardized form, such inferences can be
short-circuited in practice.
16. What kind of phenomenon is calling attention-as done by stress, panl-,
phrase, and other devices like clefting, preposing, inverted word order? Are'
these phenomena matters of meaning and therefore part of the subject of
semantics? Or are they relevant only to pragmatics?
17. Schreiber thinks this fact supports the performative analysis. However, at
best it is consistent with the performative analysis. To support that analysis, it
must be explainable in terms of higher performatives. Schreiber's data are
cases like the following:
Notes to pp. 221-232 306 Notes to pp. 232-240 307

(a) Sam
admitted
discovered
! that
possibly I
unfortunately . the earth is fiat.
24. Reported by Associated Press, Nov. 11, 1973, under the curious title:
1 1 "South African Motorists Mangle Facts in Accidents."

(a) *Sam
noted
admitted
discovered
! that
apparently
candidly
confidentially
! the earth is fiat.
Notes to Chapter 11

1noted 1 bluntly
The (a) cases are grammatical (and more or less acceptable) instances of
sentence adverbs as clause modifiers in predicate complement constructions,
whereas the adverbs of manner in the (b) cases produce, in the same en-
vironments, ungrammatical sentences. (This is Schreiber's claim-it is a
matter of considerable theoretical dispute just what sorts of unacceptability
constitute ungrammaticality; we are inclined to regard some sorts of seman-
tic nonsense as grammatical.) The performative analysis accounts for the
difference between (a) and (b) because the (b) cases would have to be de-
rived from impossible structures like (bP), which violate the rule that per-
1. Also, it does not tell us certain things we would like to know. What exactly
are these messages and how are they related to such things as the semantics of
the language and the intentions, beliefs, and desires of the communicants? How
can the picture be extended to other speech acts like promises, apologies,
verdicts, and greetings?
2. See Harnish (1977b), Dennett (1977), and Morton (1978).
3. Such a system can include a number of distinct but intertranslatable sub-
systems. See Harnish (1977b, 174) for further discussion and references.
4. Fodor (1975,28-29) gives a similar schema for the case of considered action
in general.

(bP) *Sam
1
admitted
discovered
noted
!
formative clauses cannot be freely embedded in other structures.

that I tell you

1candidly
confidentially
bluntly
I 5. Fodor, Bever, and Garrett (1974, 375) mention three kinds of factors that
that the earth is fiat.
Unfortunately, this rule is nothing more than an ad hoc device to rule out
counterexamples to the performative analysis. A more straightforward expla-
nation of the unacceptability of (b) and of (bP) would be in terms of conceptual
nonsense, not (linguistic) ungrammaticality.
18. Presumably, an account can be given of why, on the performative analysis,
(39) is not derived from *"1 request truthfully that you tell me whether you lied
to me."
19. Presumably, an account can be given on the performative analysis for why
(42) and (43) are not derived, respectively, from "I tell you fortunately that you
lied to me" and "I tell you clearly that you lied to me."
20. See Chomsky (1965, ch. 1.2) for the classical elaboration of the distinction
between grammaticality and acceptability.
21. H may not realize consciously that the locution is not a grammatical sen-
tence. Nevertheless, he can augment the locution with the parts that would
make it complete. He knows the meaning of this complete sentence and can
thereby determine what S meant to say at the locutionary level, something of
the form *( ...p ... ). In this way the SAS is unblocked.
22. There are also lexical liberties, such as using "He was stabbed by three
nameless inmates" instead of "He was stabbed by three inmates whose names
we do not know." We do not pursue lexical liberties here, though they seem to
be related to nonliteral uses of language such as metaphor.
23. What would a theory of pronominalization look like that had to cover, in ~he
same way, "He looks like he could lift a ton and throw it" and "He looks hke
he could lift a ton but not spell it."
can contribute to the determination of the PI: what the speaker is attending
to, his motivational state, and the contents of his memory. We do not dis-
pute any of these; indeed we will try to fiesh them out, for they are not very
informative.
6. The idealizations to the effect that alternatives are not always considered
and consequences computed remain in effect for SP. In many (most?) cases
Pl i = PI l and ei = el That is, the speaker's actual pragmatic intent Pl i is the
first and only one formed, and the expression uttered ei is the first and only one
constructed. In these cases most of the decision-theoretic structure of the
model is idle.
7. Osgood (1971) discusses a number of other correlations as well, concerning
pronouns, articles, and negations.
8. Fodor, Bever, and Garrett do not seem to be consistent in their description
of these conclusions. They often describe the model as working "left to right,
top to bottom" (pp. 418, 434). This makes a difference in that they also endorse
(p. 418) the claim that at least the studies by Forster (1967, 1968) "provide the
clearest evidence so far available for the theory that surface trees are elab-
orated roughly in the order that Yngve's model requires." But as Fodor, Bever,
and Garrett note earlier (p. 407), Yngve's model (and Forster's experimental
design) requires that these trees be constructed from top to bottom and left to
right in order to get structures in memory having the shape of the tree in figure
11.2. In view of this endorsement it is extremely puzzling why Fodor, Bever,
and Garrett comment (p. 419), "The Yngve model would have such bizarre
consequences as the following: since the tree is elaborated from top to bottom,
its general structure must be chosen before its lexical contents." They do not
say why, especially in light of Forster's work, this is bizarre.
9. See V. Valian (1977) for a review of some of the current theories that do not
stress these structural parameters of speech production and, because they do
not, are in various ways inadequate.

Notes to pp. 242-252 308
10. Of course there may be such asymmetries which are too small to be de-
tected by introspection. At this point the evidence would become experimen-
tal, not intuitive.
11. Fodor, Fodor, and Garrett (1975, fn. 4) acknowledge this possibility but
their alternative, which is simply to "acknowledge a class of negative primi-
tives that includes both the explicit and the implicit negatives" strikes us as
weak. It suggests, counterintuitively, that there is no direct connection be-
tween the syntactic facts involving these negatives and their semantic rep-
resentation as negatives.
12. Katz (1977c, 61) represents "bachelor" in essentially this way.
13. MacKay bases some of his conclusions on the functioning of a well-studied
amnesiac known as H.M. As a result of bilateral removal of mesial parts of the
temporal lobes and the hippocampus, H.M. is thought to be unable to form new
long-term memory traces but has normal short-term memory. According to
MacKay (1973, 37), H.M. was able to disambiguate lexical and surface am-
biguities but not underlying ambiguities; MacKay concluded, "We suggest that
hippocampal patients will be unable to learn or fully process the underlying
relations of syntactic structures they have not encountered in the past." This
explanation is suspect. If H.M. can understand normal sentences constructed
out of familiar lexical items, then he must be using his knowledge of underlying
syntactic relations just to get subject and object correct in sentences like "He is
easy (eager) to please." So if H.M. has access to this grammatical information
in comprehension, why should he not also have access to it in disambiguation?
If the answer is supposed to involve the notion of "unencountered underlying
relations of syntactic structures" then the answeris obscure at best. What is an
"unencountered" underlying relation for a mature speaker of a language? Fi-
nally, recent work on organic brain disease patients with severe ~ong-t~rm
memory dissolution shows no tendency to favor lexical and surface dIsambigu-
ation over the disambiguation of underlying ambiguities (see Bayles 1979).
-14. The probability that at least one of two independent states will be picked is
generally greater than the probability of picking one in isolation: P1 or P2 =
(P 1 + P2) - (Pl P2). However, we should he very suspicious of the assumptions
required to make the mathematics fit this case, e.g., that the "space" at hand
needs to be searched for a reading. If both readings are being computed, then
both readings are available to the hearer.
15. Another part of the problem of identifying the type of clausal unit has to do
with the notion of a main clause. For example, many grammars give the sen-
tence "John wanted to leave" the underlying structure:
[Sl [NPI JohnNPJ [VPI [v wanted v] [sz [NPz John NPz] [vPz[v leave v] vpz] sz] vP I] sJ
The surface form of the sentence results when NP2 is deleted under identity
with NP1. We can see that "John leave" in S2 might be construed as a clause.
But how about" John wanted" in Sl? Is this an underlying clause? Or is the
whole complex "John wanted John leave" the main underlying clause? We
have assumed in the text that each underlying occurrence of S defines an
Notes to pp. 252-278 309
underlying clause. If that position is not adopted and instead one says that
subordinate S nodes (or labeled brackets) are part of one single underlying
clause, then it is possible for the underlying clause to be the unit of comprehen-
sion. Indeed, Carroll and Bever (1976, 324) suggest that this latter position is
the only one when they write, "We have suggested that the deep sentoid, and
therefore its surface realization, the clause, is the primary unit of sentence
perception" (emphasis added). So it seems that the matter is both empirical and
terminological.
We have ignored, so far, a purported third alternative to these views, namely,
the position of J. Carroll et al. (see for instance Carroll 1978), wherein the
relevant notion of a clause is said to be "functional" as well as structural.
However, on inspection it turns out that the notion offunction at work is, from
the point of view oflanguage use, pretty thin: "We refer to linguistic sequences
which provide the listener with an intact subject-verb-(object) group as func-
tionally complete sequences, and we predict that such sequences will be 'good'
sentence perception units" (p. 507). From a speech act point of view this looks
like a version of the surface structure conception tentatively endorsed in the
text, restricted to certain forms.
Notes to Appendix
1. Lewis (1969, 52ft') and Schiffer (1972, 30ff) use the terms common knowledge
and mutual knowledge, respectively, which needlessly imply the truth of what
is mutually believed. Their definitions are not limited to three levels of belief
but go on indefinitely. Higher-level beliefs are in principle possible, and indeed
among spies or deceptive intimates there could be divergence at level four or
higher without divergence at the first three levels, but we think such higher-
level beliefs are not possible for a whole community or large group. Never-
theless, to allow for this possibility our definition could be amended to require
that no higher-level belief, if there are any, be false.
2. A may be any of a wide range of behaviors rather than one particular kind,
which have some feature in common. We might speak of A (in C) as the range
of a norm and G, the collectivity to which it applies, as its scope. Also, note
that in section 7.1 we used DF2 to define rule, but here we reserve that term for
the special cases covered by DFg-DF ll .
3. In general when "social" modifies any of our defined expressions, it implies
'socially real by virtue of mutual belief.'
4. In effect each type of rule to be defined contains a special reason that
amplifies the should clause (iii) in our definition of social norms.
5. A subcategory of a role category may fail to be a role category. Subcate-
gories like 24-year-olds or 158~pounders are too specific to have their own
regularities or standards.
6. The concept of social class is a special case of social position. In speaking of
classes, whether economic, political, racial, or otherwise, we imply that the
Notes to pp. 278-282 310

society in question is as a whole divisible into classes, such that each person is a
member of one and only one class. Of course there may be borderline cases
Bibliography
where the society does not divide up neatly or where there is social mobility.
Furthermore, the social stratification of classes is associated with a value scale
reckoned by power, wealth, or prestige. Accordingly, there are norms for the
behavior and for the treatment of members of each class.
7. Even in social scientific contexts, the term group is sometimes used so
broadly as to include role categories or even categories simpliciter, as in ethnic
group, age group, opinion group, and reference group. Members of such groups.
need not share any degree of structured relationship, although they could form
groups in the narrower sense of social group defined in DF 16
8. Group identification (and alienation) are discussed in Bach 1973, ch. 2.

Akmajian, A., .R. Demers, and R. Harnish. 1979. Linguistics: An Introduction

to Language and Communication. Cambridge, Mass.: MIT Press.
Akmajian, A., and F. Heny. 1975. An Introduction to the Principles of Trans-
formational Syntax. Cambridge, Mass.: MIT Press.
Anderson, S. 1971. "On the linguistic status of the performative-constative
distinction." Bloomington: Indiana University Linguistics Club.
Austin, J. L. 1962. How to Do Things with Words. Cambridge, Mass.: Harvard
University Press.
Bach, E. 1970. "Problominalization." Linguistic Inquiry 1: 121-122.
Bach, K. 1970. "Part of what a picture is." British Journal of Aesthetics
10: 119-137.
Bach, K. 1973. Exit-Existentialism. Belmont, Calif.: Wadsworth.
Bach, K. 1975a. "Performatives are statements too." Philosophical Studies
28:229-236.
Bach, K. 1975b. "Analytic social philosophy-basic concepts." Journalfor the
Theory of Social Behavior 5:189-214.
Bach, K. 1977. "When to ask, 'What if everyone did that?''' Philosophy and
Phenomenological Research 37:464-481.
Bach, K. 1978. "A representational theory of action." Philosophical Studies
34:361-379.
Bates, E. 1976. Language and Context: The Acquisition of Pragmatics. New
York: Academic Press.
Bayles, K. 1979. Language Profiles of a Geriatric Population, Doctoral disser-
tation, University of Arizona.
Bennett, J. 1973. "The meaning nominalist strategy." Foundations of Lan-
guage 10: 141-168.
Bennett, J. 1976. Linguistic Behavior. Cambridge, England: Cambridge Uni-
versity Press.
Bibliography 312
Bever, T., M. Garrett, and R. Hurtig. 1973. "The interaction of perceptual
processes and ambiguous sentences." Memory and Cognition 1:277-286.
Bever, T., M. Garrett, and R. Hurtig. 1976. "Projection mechanisms in read-
ing." Journal of Psycholinguistic Research 2:215-226.
Bever, T., J. Katz, and T. Langendoen, eds. 1976. An Integrated Theory of
Linguistic Ability. New York: Crowell.
Bierwisch, M. 1970. "Semantics." In J. Lyons, ed., New Horizons in Linguis-
tics. Baltimore: Penguin.
Boer, S., and W. Lycan. 1976. "The myth of semantic presupposition."
Bloomington: Indiana University Linguistics Club.
Brewer, W. 1977. "Memory for the pragmatic implications of sentences."
Memory and Cognition 6:673-678.
Brewer, W., R. Harris, and E. Brewer. "Comprehension of literal and figura-
tive meaning." Unpublished manuscript, n.d. University of Illinois.
Cairns, H., and C. Cairns. 1976. Psycholinguistics. New York: Holt, Rinehart
& Winston.
Carey, P. W., J. Mehler, and T. Bever. 1970. "When do we compute all the
interpretations of an ambiguous sentence?" Advances in Psycholinguistics,
61-75.
Carroll, J. 1978. "Sentence perception units and levels of syntactic structure."
Perception and Psychophysics 6:506-514.
Carroll, J., and T. Bever. 1976. "Sentence comprehension: A case study in the
relation of knowledge and perception." In E. Carterette and M. Friedman,
eds., Handbook of Perception, vol. 7. New York: Academic Press.
Castaneda, H.-N. 1977. "On the philosophical foundations of the theory of
communication: Reference." Midwest Studies in Philosophy 2: 165 -186.
Chomsky, N. 1957. Syntactic Structures. The Hague: Mouton.
Chomsky, N. 1965. Aspects of the Theory of Syntax. Cambridge, Mass.: MIT
Press.
Chomsky, N. 1977. Essays on Form and Interpretation. Amsterdam: Elsevier
N orth-Holland.
Clark, H., and E. Clark. 1977. Psychology and Language. New York: Harcourt
Brace Jovanovich.
Clark, H., and S. Haviland. 1974. "Psychological processes as linguistic ex-
planation." In D. Cohen, ed., Explaining Linguistic Phenomena. New York:
Hemisphere Publishing Co.
Clark, H., and S. Haviland. 1977. "Comprehension and the given-new con-
tract." In Freedle.
Clark, H., and P. Lucy. 1975. "Inferring what was meant from what was said."
Journal of Verbal Learning and Verbal Behavior 14:56-72.
Bibliography 313
Cohen, L. J. 1974. "Speech Acts." In T. Sebeok, ed., Current Trends in Lin-
guistics. The Hague: Mouton.
Cole, P., and J. Morgan, eds. 1975. Syntax and Semantics, vol. 3. New York:
Academic Press.
Davidson, D. 1967. "Truth and meaning." Synthese 17:304-323.
Davidson, D. 1970. "Semantics of natural languages." Reprinted in Davidson
and Harman, eds., The Logic of Grammar. Encino, Calif.: Dickenson, 1975.
Davidson, D., and G. Harman, eds. 1972. Semantics of Natural Language.
Dordrecht, Holland: Reidel.
Davison, A. 1975. "Indirect speech acts and what to do with them. In Cole and
Morgan, eds., Syntax and Semantics, vol. 3.
Dennett, D. 1977. Review of J. A. Fodor, The Language of Thought. Mind
86:265-280.
Donnellan, K. 1966. "Reference and definite descriptions." Philosophical Re-
view 75:281-304.
Evans, G., and J. McDowell, eds. 1976. Truth and Meaning. Oxford, England:
Oxford University Press.
Field, H. 1978. "Mental representation." Erkenntnis 13:9-61.
Fillmore, C. 1971. "Verbs of judging." In Fillmore and Langendoen, eds.,
Studies in Linguistic Semantics. New York: Holt, Rinehart & Winston.
Fillmore, C., and T. Langendoen, eds. 1971. Studies in Linguistic Semantics.
New York: Holt, Rinehart & Winston.
Fodor, J. A. 1975. The Language of Thought. New York: Crowell.
Fodor, J. A., T. Bever, and M. Garrett. 1974. The Psychology of Language.
New York: McGraw-Hill.
Fodor, J. D. 1977. Semantics: Theories of Meaning in Generative Grammar.
New York: Crowell.
Fodor, J. D., J. A. Fodor, and M. Garrett. 1975. "The psychological unreality
of semantic representation." Linguistic Inquiry 6:515-531.
Forster, K. 1967. "Sentence completion latencies as a function of constituent
structure." Journal of Verbal Learning and Verbal Behavior 6:878-883.
Forster, K. 1968. "Sentence completion in left-and-right branching languages."
Journal of Verbal Learning and Verbal Behavior 7:296-299.
Foss, D. J., T. Bever, and M. Silver. 1968. "The comprehension and verifica-
tion of ambiguous sentences." Perception and Psychophysics 4:304-306.
Fraser, B. 1971. "An analysis of 'even'." In Fillmore and Langendoen, eds.,
Studies in Linguistic Semantics, New York: Holt, Rinehart & Winston.
Fraser, B. 1973. "A partial analysis of vernacular performative verbs." Austin:
Texas Performadillo Conference.
Bibliography 314
Fraser, B. 1975. "Hedged performatives." In Cole and Morgan, eds., Syntax
and Semantics, vol. 3.
Freedle, R., ed. 1977. Discourse Production and Comprehension, vol. 1. Nor-
wood, N.J.: Ablex Publishing Co.
Frege, G. 1892. "On sense and reference." In P. Geach and M. Black, eds.,
Translations From the Philosophical Writings ofG. Frege, Oxford: Blackwell.
Frye, M. 1973. "Force and meaning." Journal of Philosophy 70:281-294.
Garrett, M. 1970. "Does ambiguity complicate the perception of sentences?"
Advances in Psycholinguistics, 48-60.
Ginet, S., and C. Ginet. 1976. Review ofZ. Vendler, Res Cogitans. Philosoph-
ical Review 85:216-224.
Goldman, A. 1970. A Theory of Human Action. Englewood Cliffs, N.J.:
Prentice-Hall.
Green, G. 1975. "How to get people to do things with words." In Cole and
Morgan, eds., Syntax and Semantics, vol. 3.
Greenbaum, S. 1969. Studies in English Adverbial Usage. Miami, Fla.: Miami
Linguistics Series.
Grice, H. P. 1957. "Meaning." Philosophical Review 66:377-388.
Grice, H. P. 1961. "The causal theory of perception." Proceedings of the
Aristotelian Society, Supplementary Volume 35:121-152.
Grice, H. P. 1967. William James Lectures. Unpublished.
Grice, H. P. 1968. "Utterer's meaning, sentence-meaning and word-meaning."
Foundations of Language 4:225-242.
Grice, H. P. 1969. "Utterer's meaning and intentions." Philosophical Review
78:147-177.
Grice, H. P. 1975. "Logic and conversation." In Cole and Morgan, eds., Syn-
tax and Semantics, vol. 3.
Hall, R. 1959. "Excluders." Analysis 20:1-7.
Harman, G. 1973. Thought. Princeton, N.J.: Princeton University Press.
Hamish, R. 1972. Studies in Logic and Language. Doctoral dissertation, Mas-
sachusetts Institute of Technology.
Harnish, R. 1975. "The argument from lurk." Linguistic Inquiry 6:145-154.
Reprinted in Bever, Katz, and Langendoen, eds., An Integrated Theory of
Linguistic Ability, N ew York: Crowell.
Harnish, R. 1976a. "Two consequences of transparent subject position." Philo-
sophical Studies 30: 11-18.
Hamish, R. 1976b. "Logical form and implicature." In Bever, Katz, and
Langendoen, eds., An Integrated Theory of Linguistic Ability, New York:
Crowell.
Bibliography 315
Harnish, R. 1977. Review of M. Dummett, Frege: Philosophy of Language,
1973. General Linguistics 17:47-58.
Harnish, R. 1977a. "Searle on Katz's semantic theory." The Southwestern
Journal of Philosophy 8:23-32.
Harnish, R. 1977b. Review of J. A. Fodor, -The Language of Thought. The
Journal of Psycholinguistic Research 6: 172-184.
Harnish, R. 1978. Review of J. Sadock, Toward a Linguistic Theory of Speech
Acts. Lingua 44:288-301.
Harnish, R. "A projection problem for pragmatics." To appear in F. Heny and
H. Schnelle, eds., Syntax and Semantics, vol. 10. New York: Academic Press
(in press).
Hamish, R. Fregean Semantics (in preparation).
Harris, R. 1974. "Memory and comprehension of implications and inferences
of complex sentences." Journal of Verbal Learning and Verbal Behavior 13:
626-637.
Harris, R., and G. Monaco. 1978. "Psychology of pragmatic implication: In-
formation processing between the lines." Journal of Experimental Psychology
107:1-22.
Hart, H. L. A. 1961. The Concept of Law. Oxford, England: Oxford University
Press.
Haviland, S., and H. Clark. 1974. "What's new? Acquiring new information as
a process in comprehension." Journal of Verbal Learning and Verbal Behavior
13:512-521.
Heringer, J. 1972. Some Grammatical Correlates of Felicity Conditions and
Presupposition. Bloomington: Indiana University Linguistics Club.
Holmes, V., R. Arwas, and M. Garrett. 1977. "Prior context and the percep-
tion of lexically ambiguous sentences." Memory and Cognition 5:103-110.
Hungerland, 1. 1960. "Contextual implication." Inquiry 3:211-258.
Hutchinson, L. 1971. "Presupposition and belief-inferences." Chicago Lin-
guistic Society 7.
Jackendoff, R. 1972. Semantic Interpretation in Generative Grammar. Cam-
bridge, Mass.: MIT Press.
Just, M., and H. Clark. 1973. "Drawing inferences from the presuppositions
and implications of affirmative and negative sentences." Journal of Verbal
Learning and Verbal Behavior 12:21-31.
Karttunen, L. 1973. "Presuppositions of compound sentences." Linguistic In-
quiry 4:169-193.
Katz, J. J. 1966. The Philosophy of Language. New York: Harper & Row.
Katz, J. J. 1972. Semantic Theory. New York: Harper & Row.
Bibliography 316
Katz, J. J. 1975. "Logic and language." In K. Gunderson, ed., Language, Mind
and Knowledge. Minneapolis: University of Minnesota Press.
Katz, J. J. 1977a. "A proper theory of names." Philosophical Studies 31: 1-80.
Katz, J. J. 1977b. "The real status of semantic representations." Linguistic
Inquiry 8:559-584.
Katz, J. J. 1977c. Propositional Structure and Illocutionary Force. New York:
Crowell.
Katz, J. J., and J. A. Fodor. 1963. "The structure of a semantic theory."
Language 39: 170-210.
Katz, J. J., and T. Langendoen. 1976. "Pragmatics and presupposition." Lan-
guage. In Bever, Katz, and Langendoen, eds., An Integrated Theory of Lin-
guistic Ability, New York: Crowell.
Keenan, E. 1971. "Two kinds of presupposition in natural language." In
Fillmore and Langendoen, eds. Studies in Linguistic Semantics, New York:
Holt, Rinehart & Winston.
Keenan, E. 1976. "On the universality of conversational implicatures." Lan-
guage in Society 5:68-81.
Kempson, R. 1975. Presupposition and the Delimitation of Semantics. Cam-
bridge, England: Cambridge University Press.
Klapp, O. E. 1957. "The concept of consensus and its importance." Sociology
and Social Research 41:336-342.
Kripke, S. 1972. "Naming and necessity." In Davidson and Harman, eds.,
Semantics of Natural Language, New York: Humanities Press.
Kripke, S. 1977. "Speaker's reference and semantic reference." Midwest
Studies in Philosophy 2:255-276.
Lackner, J. R., and M. Garrett. 1973. "Resolving ambiguity: EtIects of biasing
context in the unattended ear." Cognition 1:359-372.
Laing, R. D. 1968. The Politics of Experience. New York: Ballantine Books.
LakotI, G. 1970. "Linguistics and natural logic." Synthese 22: 151-271.
LakotI, G. 1972. "Performative antimonies." Foundations of Language 8:
569-572.
LakotI, G. 1975. "Pragmatics and natural logic." In E. Keenan, ed., Formal
Semantics of Natural Language. Cambridge, England: Cambridge University
Press.
Landis, J. R. 1971. Sociology: Concepts and Characteristics. Belmont, Calif.:
Wadsworth.
Lemmon, E. 1962. "Sentences verifiable by their use." Analysis 12:86-89.
Lewis, D. 1969. Convention. Cambridge, Mass.: Harvard University Press.
Lewis, D. 1970. "General semantics." Synthese 22:18-67.
Bibliography 317
Loar, B. 1976. "The semantics of singular terms." Philosophical Studies 30:
353-378.
Locke, J. 1691. An Essay Concerning Human Understanding.
MacKay, D. 1966. "To end ambiguous sentences." Perception and Psycho-
physics 1:426-436.
MacKay, D. 1973. "Aspects of the theory of comprehension, memory, and
attention." The Quarterly Journal of Experimental Psychology 25:22-40.
MacKay, D., and T. Bever. 1967. "In search of ambiguity." Perception and
Psychophysics 2: 193-200.
Marslen-Wilson, W. 1975. "Sentence perception as an interactive parallel pro-
cess." Science 189:226-228.
McCawley, J. 1973. "Remarks on the lexicography of performative verbs."
Austin: Texas Performadillo Conference.
Miller, G. A., and Johnson-Laird, P. N. 1976. Language and Perception. Cam-
brige, Mass.: Harvard University Press.
Montague, R. 1974. Formal Philosophy. Edited by R. Thomason. New Haven:
Yale University Press.
Morton, A. 1978. Review of J. A. Fodor, Language of Thought. Journal of
Philosophy 75: 161-169.
Muraki, M. 1972. "Discourse presupposition." In Papers in Linguistics 5 .2.
Oden, G. 1978. "Semantic constraints and judged preference for interpreta-
tions of ambiguous sentences." Memory and Cognition 1:26-37.
Oden, G., and J. Spira. 1978. "Influence of context on the activation and
selection of ambiguous w<;>rd senses." Wisconsin Human Information Process-
ing Program, report number 6.
Osgood, C. 1971. "Where do sentences come from?" In Steinberg and Jakobo-
vits, eds., Semantics, Cambridge, England: Cambridge University Press.
Partee, B. 1977. "Possible worlds semantics and linguistic theory." The Monist
60:303-326.
Quine, W. 1960. Word and Object. Cambridge, Mass.: MIT Press.
Rawls, J. 1955. "Two concepts of rules." Philosophical Review 64:3-32.
Rosenberg, J. 1974. Linguistic Representation. Dordrecht, Holland: Reidel.
Ross, Sir D. 1930. The Right and the Good. London, England: Oxford Univer-
sity Press.
Ross, J. 1970. "On declarative sentences." In R. Jacobs and P. Rosenbaum,
eds., Readings in English Transformational Grammar. Lexington, Mass.: Ginn.
Ross, J. 1974. "Three batons for cognitive psychology." In W. Weimer and D.
Palermo, eds., Cognition and the Symbolic Process. Hillsdale, N.J.: Earlbaum.
Bibliography 318
Russell, B. 1905. "On denoting." Mind 14:479-493.
Sadock, J. 1972. "Speech act idioms." Chicago Linguistic Society 8.
Sadock, J. 1974. Toward a Linguistic Theory of Speech Acts. New York:
Academic Press.
Scheff, T. 1967. "Toward a sociological model of consensus." American Socio-
logical Review 32:32-46.
Schelling, T. 1960. The Strategy of Conflict. Cambridge, Mass.: Harvard Uni-
versity Press.
Schiffer, S. 1972. Meaning. London, England: Oxford University Press.
Schreiber, P. 1972. "Style disjuncts and the performative analysis." Linguistic
Inquiry 3:321-347.
Schvaneveldt, R., D. Meyer, and C. Becker. 1976. "Lexical ambiguity, seman-
tic context, and visual word recognition." Journal of Experimental Psychology:
Human Perception and Performance 2:243-256.
Schweller, K., W. Brewer, and D. Dahl. 1976. "Memory for illocutionary
forces and perlocutionary effects of utterances." Journal of Verbal Learning
and Verbal Behavior 15:325-337.
Searle, J. 1964. "How to derive 'ought' from 'is'." Philosophical Review 73:
43-58.
Searle, J. 1965. "What is a speech act?" Reprinted in J. Searle, ed., The
Philosophy of Language. London, England: Oxford University Press, 1971.
Searle, J. 1968. "Austin on locutionary and illocutionary acts." Philosophical
Review 77:405-424.
Searle, J. 1969. Speech Acts. Cambridge, England: Cambridge University
Press.
Searle, J. 1975a. "Indirect speech acts." In Cole and Morgan, eds., Syntax and
Semantics, vol. 3.
Searle, J. 1975b. "A taxonomy of illocutionary acts." In K. Gunderson, ed.,
Language, Mind and Knowledge. Minneapolis: University of Minnesota Press.
Secord, P., and C. Backman. 1974. Social Psychology, 2nd ed. New York:
McGraw-Hill.
Singer, M. 1976. "Context inferences in the comprehension of sentences."
Canadian Journal of Psychology 30:39-46.
Steinberg, D., and Jakobovits, L., eds. 1971. Semantics. Cambridge, England:
Cambridge University Press.
Stenius, E. 1967. "Mood and language game." Synthese 17:254-274.
Strawson, P. 1950. "On referring." Mind 59:320-344.
Strawson, P. 1952. Introduction to Logical Theory. London: Methuen.
Strawson, P. 1964a. "Intention and convention in speech acts." Philosophical
Review 73:439-460.
Bibliography 319
Strawson, P. 1964b. "Identifying reference and truth values." Theoria 30:96-
118.
Strawson, P. 1971. "Meaning and truth." In Logic-Linguistic Papers. London:
Methuen.
Suppes, P. 1957. Introduction to Logic. New York: Van Nostrand Reinhold.
Tannenbaum, P., and F. Williams. 1968. "Generation of active and passive
sentences as a function of subject or object focus." Journal of Verbal Learning
and Verbal Behavior 7:246-250.
Thomson, J. 1977. Acts and Other Events. Ithaca, N.Y.: Cornell University
Press.
Tyler, L., and W. Marslen-Wilson. 1977. "The on-line effects of semantic
context on semantic processing." Journal of Verbal Learning and Verbal Be-
havior 16:683-692.
Urmson, J. 1952. "Parenthetical verbs." Reprinted in C. Caton, ed., Philos-
ophy and Ordinary Language. Urbana: University of Illinois Press.
Valian, V. 1977. "Talk, talk, talk: A selective critical review of theories of
speech production." In Freedle, ed., Discourse Production and Comprehen-
sion, Norwood, N.J.: Ablex Publishing Co.
Vendler, Z. 1972. Res Cogitans. Ithaca, N.Y.: Cornell University Press.
Weiser, A. 1974. "Deliberate ambiguity." Chicago Linguistic Society 10,
721-730.
Weiser, A. 1975. "How to not answer a question: Purposive devices in conver-
sational strategy." Chicago Linguistic Society 11.
Wilson, D. 1975. Presupposition and Non-Truth-Conditional Semantics. New
York: Academic Press.
Ziff, P. 1967. "On H. P. Grice's account of meaning." Analysis 28: 1-8.

Acknowledgments, 40, 41, 51-55,
63, 64, 81, 103, 122, 125-126,
288n9
Advisories, 39-40, 48, 49,
289n3
Akmajian, A., 140
Ambiguity, 5, 9, 20, 21, 22, 80,
101-102, 107, 133-134, 138, 140,
144, 150, 164, 224, 245-247,
295n5, 307n13. See also
Disambiguation
comprehension of, 245-247
versus conventionality, 191-192
Ambiguity thesis, 174-183,
302n17
arguments against, 184-187
Analyticity, 143
Answers, 40, 63
Apologize, 40, 51, 53, 54, 72, 75
Arwas, R., 249
Ascriptives, 42, 44
As-if facts, 115
Aspectuals, 162
Assertives, 36, 39, 42-46, 56
Austin, J. L., 4, 13, 16-17, 19, 20,
38, 40, 47, 49, 51, 55, 56, 120,
131, 132, 203, 204, 205, 206,
285nl, 285n3, 288nll, 289nl,
304n2, 304n5, 304n8, 305nl0
Bach, E., 302n16
Bach, K., 275, 310n8
Backman, C., 271, 277, 278
Bates, E., 159
Becker, C., 253 - 254
Behabitives, 41, 51
Bennett, J., 286
Index
Bever, T., 141, 144, 145, 234, 240,
245, 247, 249, 251, 252, 253, 254,
257, 296n5, 307n5, 307n8, 309n15
Bidding, 36, 51, 52-53, 54
Bierswich, M., 138, 139, 141
Boer, S., 158
Bracketing (Br), 162-164
Brewer, W., 258, 261, 262, 265, 266
Can, 196-197
Carey, P. W., 253
Carroll, J., 252, 309n15
Casteneda, R., 287n5
Changing the subject, 99, 105-106
Chomsky, N., 140, 147,230, 306n20
Circumlocution, 99
Clark, E., 240
Clark, R., 257-260, 263, 265
Clausal-closure hypothesis, 245-247,
248-257
Collateral acts, 96, 97-103
Commissives, 36, 37, 40, 41, 49-51,
64, 81, 125
Communicative intention, 1-16,
39-59, 232, 237, 261. See also
Illocutionary intentions
Communicative presumption (CP), 7,
10 12-15, 34, 58-59, 60-61, 65,
78,80,84,87,88,89,96,97,99,
116, 208, 213, 214, 225, 230,
292nl, 292n4, 293n5
Communicative success, 3-4, 5,
13-16, 39, 57, 86-90, 110
Compatibility condition (CC), 34-35
Competence, linguistic, 9, 89, 230
Concessives, 36, 43, 45
Confirmatives, 36, 39, 42-43, 46
Index
Congratulate, 36, 51, 52, 54, 79
Consensus, 268, 269
Constatives, 36, 37, 40, 41, 42-47,
54, 63, 71, 81, 115, 123-124, 170,
212,289n3
performatives as, 203-208
Constitutive rules, 127-131, 289n4,
293nl, 295n3
Contextual appropriateness, 21, 22,
62,65
Contextual implication, 166
Contextual relevance, 45
Contextual selection, 21, 247-257
Convention, 108-110
role in communication of, 120-135
passim
Conventionality thesis, 183-192
Conventions of use, 188
Conversational appropriateness, 62,
65
Conversational implicature, 165-172
Conversational paraphrase, 227, 228
Conversational presumptions, 62-63,
65, 70, 72
Cooperative principle, 166-167
Could, 196-197
Counts-as rules, 108, 125
Covert collateral acts, 97, 101-102
Crowds, 277, 280
Dahl, D., 261, 265-266
Davidson, D., 70, 138, 139, 295n3
Declarative (sentences), 9, 25, 35,
297nll
Definite descriptions, 25-27, 156,
165,287n5
Demers, R., 140
Deniability, 101
Dennett, D., 307n2
Denotation, 20
Descriptive content, 31
Descriptives, 42, 44, 45
Dialect, 172
Directives, 36, 40, 41, 47-49, 63-64,
81, 124-125
Direct literal strategy (DLS), 77
Direct nonliteral strategy (DNS), 78
Disambiguation, 23, 247-257. See
also Ambiguity
Disputative, 43, 45, 46
Dissentive, 43, 45, 46
Donnellan, K., 287n5
322
Double entendre, 102
Durkheim, E., 284
Effectives, 40, 108, 110-115
Entailment, 143, 144, 145, 149, 158,
161
Essential condition, 56, 128, 131
Essential rule, 129
Evans, G., 138
Exaggeration, 66, 67, 68, 69
Exercitive, 41, 47
Explicit performatives, 131-132,
203-209, 304nl, 304n2
Expositives, 40
Expressing (attitudes), 15-16, 39-41
Facetiousness, 67-68
Factives, 164-165
False consensus, 268
False descriptions, 30-31
Falsehood, obvious factual, 68
Felicity conditions, 55-57
Field, H., 245
Figure of speech, 68
Fillmore, C., 155, 159, 165
Fodor, J. A., 138, 139, 141, 142,
143, 144, 145, 148, 236, 239, 240,
242-243, 244, 254, 292n3, 296n5,
307n4, 307n5, 307n8, 308nll
Fodor, J. D., 140, 141, 142,
242-243, 244, 308nl1
Force-determinacy (F-determinacy),
36, 173
Foss, D. J., 253
Fraser, B., 209, 210, 211-212,
213-214, 215, 216-218, 289nl,
305n13, 305n14
Frege, G., 155, 156, 297n13, 298n15
Frye, M., 150
Garrett, M., 141, 142, 144, 145, 236,
240, 241-242, 244, 247, 248, 249,
250, 251, 252, 254, 257, 296n5,
307n5, 307n8, 308nl1
Generalized conversational
implicature, 170-172
Generative semantics, 139
Ginet, C., 304n4
Ginet, S., 304n4
Given-New strategy, 257-258, 260
Goldman, A., 285n2
Grammatical form, 145, 146, 148
Index
Grammatical inference rules, 147
Gramm atic ality , 229, 230
Green, G., 197, 301n8
Greenbaum, S., 222
Greeting, 51, 52, 63, 294n2
Grice, H. P., 13-14,62-63, 64, 103,
137, 150, 151, 152, 153, 154, 161,
165-166, 167-168, 169-170, 171,
172, 213, 214, 215, 286n12,
286n13, 291nl, 291n5, 293n5,
295n6, 296n9, 299n25, 299n26,
299n27, 299n28, 300n29
Gricean maxims, 104-107
Groups, 280- 282
Guarantee, 50
Hall, R., 291n5
Harman, G., 138, 293n7
Harnish, R. M., 65, 140, 144, 148,
164, 168, 171, 176, 291nl,
298n13, 299n24, 300n34, 304n6,
307n2
Harris, R., 260, 262
Hart, H. L. A., 283
Haviland, S., 257-258
Hedged performatives. See
Performatives, hedged
Heny, F., 140
Heringer, J., 56, 70
Higher performative analysis,
220-221, 224-227, 303n24
Holmes, V., 249
Hungerland, I., 166
Hurtig, R., 247, 249, 250, 251, 252,
257
Hutchinson, L., 102
Hyperbole, 66
Idiolect, 172
Idioms, 186, 188, 232
Illocutionary acts" 3-4, 13, 34,
285n3, 288nl1, 289n4, 289n6
classification of, 41, 42- 55,
110-113
and communicative conventions,
122-127
conventional, 108-119 passim
and institutional facts,
113-116
performance of, 116-119
indirect, 56, 70-80, 173-202
passim, 263-265, 290n6
323
literal and direct, 10-12, 22, 34,
60-65, 70, 76
nonliteral, 22, 65-70, 77, 97-98,
262-263
types of communicative, 35, 38,
39-59 passim
Illocutionary adverbials, 33-34,
219-228
Illocutionary convention (IC),
189-190
Illocutionary devices, 203-233
passim
Illocutionary effect. See Uptake
Illocutionary force, 6, 10-11, 30, 63,
192
Illocutionary force potential, 34-37,
288nll
Illocutionary intentions, 7, 12-16,
39,40,65,72,75,81,84-89,
99-100, 196-198. See also
R-intentions
hearer's identification of, 90-91
Illocutionary standardization,
173 - 202 passim
and illocutionary devices, 203-233
passim
and paragrammatic facts, 198-202
Imperative (sentences), 9, 25, 71,
80, 186, 187-188, 199, 200,
297n11
Impositives, 176-177, 180-183, 187
Indefinite reference, 134
Indexicals, 27-28, 220
Indicator of illocutionary force, 129
Indirect quotation, 9, 29-34, 286
Infelicity, 57-59, 195-198
Inference. See Speech Act Schema
(SAS)
Inference rules, 142
Inference to a plausible explanation,
92
Informing, 36, 42, 45
Infringement, 167
Innuendo, 101
Input-output conditions, 128, 131
Input-output rule, 130
Insincerity. See Infelicity
Institutional facts, 113-114
Interrogative (sentences), 9-10,
70-71,80,188,189, 199,200,
303
Intonation, 36, 288nlO
Index
Intonation contours, 67-68
sarcastic, 33
Introspection, 92-93
Inviting, 36, 51
Irony, 67-68
Jackendoff, R., 155, 159
Johnson-Laird, P., 138, 286n5
Joking, 18, 97, 98
Just, M., 258-260
Karttunen, L., 158
Katz, J. J., 138, 139, 140, 141,
142, 155, 157, 243, 287n7,
288n11, 289n1, 304n4, 305n9,
308n12
Keenan, E., 159, 298n18, 299n28
Kempson, R., 65, 155, 158, 171,
291n2, 295n1
Kidding, 97
Klapp, O. E., 268
Knowledge of language, 8
Kripke, S., 287n7
Lackner, J. R., 247, 248, 249, 250
Laing, R. D., 269
Lakoff, G., 142, 155, 158
Landis, J. R., 271, 280
Langendoen, T., 155
Lemmon, E., 305n11
Lewis, D., 138, 309n1
Lexical decomposition, 139-144,
242-245
Lexical omission, 231, 232
Lexical representation during
comprehension, 242-245
Linguistic presumption (LP), 7, 8,
20, 21, 22, 24, 60, 78, 84, 88,
116, 133, 292n1
Literal condition, 129, 131
Literal illocutionary acts. See
Illocutionary acts
Literally-based indirect strategy
(LIS), 78, 79
Literalness, presumption of (PL) , 12,
61, 78, 224, 263
Literal performance (Lit) of an
illocutionary act, 35
Literal rule, 130
Literal strategy, 79
Loar, B., 287n7
Locke, John, 234
324
Locutionary acts, 3, 4, 8, 9, 19-38
passim, 89, 134, 137-172 passim,
285n1, 288n10, 288n11
Austin's conception of, 19-20
and conventions, 132-134
hearer's determination of, 24-28
Locutionary -compatibili ty
(L-compatible), 11, 34, 36, 173
Locutionary strategy (LS), 77, 78
Locution-specific illocutionary act,
110
Logical form, 144, 145, 146, 147, 148
Lucy, P., 263, 265
Lycan, W., 158
McCawley, J., 289n1
McDowell, J., 138
MacKay, D., 245,246,247,249,
251,308n13
Malapropism, 86, 134, 231
Manner, presumption of (MA) , 64,
107, 167
Manner adverbs, 221-225
Marsten-Wilson, W., 234, 244, 254
Maxim of sufficient reason, 215
Maxims. See Gricean maxims
Meaning
attribution, 172
decompositional theories of,
139-143
inferential theories of, 143-149
intended, 9
linguistic, 8, 9, 138-149
speaker, 149-154
critique of, 151-154
varieties of, 150-151
specification, 148, 172
Meaningfulness, 140, 149
Metaphor, 68, 69, 130, 292n7,
306n22
Metaphor-potential, 69
Metonymy, 68, 292n7
Mehler, J., 253
Meyer, D., 253-254
Miller, G., 138, 286n5
Mimicking, 97, 98
Misleadingness of grammatical form,
147, 148
Mistaken factual beliefs, 33
Mistaken linguistic beliefs, 32-33
Modals, 211-212
Monaco, G., 260
Index
Montaque, R., 138
Morality, presumption of (MO), 64,
65
Muraki, M., 155
Must, 164, 198
Mutual beliefs, 109, 113, 114, 119,
121, 124, 125, 133, 184, 192, 193,
194, 208, 267-284 passim
Mutual contextual beliefs (MCBs), 5,
6, 7, 21, 25, 62-63, 66, 70, 71,
72,78,80,84,86,90,91,92-93,
94, 97, 103, 116, 300n33
Natural facts, 113
Negative factives, 162, 163
Nonliterally-based indirect strategy
(NIS),78
Normative roles, 279
Norms, 270
social, 271-273
Oden, G., 254-255
Offers, 49, 50, 79
Operative meaning, 5, 9, 20-23, 33,
36, 137, 138, 150, 157
Operative meaning stage, 241
Ordering, 36, 76, 79
Organizations, 277, 280, 282-283
Osgood, C., 238, 239, 307n7
Ought, 198
Overstatement, 66, 67
Overt acts, 97
Paragrammatic facts, 198-202
Partee, B., 138
Perceived consensus, 268
Performatives, 11, 36, 55, 131-132
as constatives, 203-208
embedded,209-211
explicit, 203-208
hedged, 209-219, 305n13, 305n14
locutionary, 209
Perlocutionary acts, 3, 4, 16-18, 81
versus illocutionary acts, 81-82
and the SAS, 82-83
Perlocutionary effect, 4, 14, 16, 39,
54
Perlocutionary intentions, 12-13,
41-44,48, 49, 50, 54, 81-82, 97,
237,293n6
Permissives, 47,49
Phatic act, 19, 20, 285n1
325
Phonetic act, 19, 285n1
Phrase-markers, 146, 148
Phrase structure grammars, 140
Please, 198-199, 200, 201-202,
303n23, 303n24, 303n25, 303n26,
304n27
Pluralistic ignorance, 268
Politeness, presumption of (PO), 64,
100, 106, 196
Position, 278, 279
Possibly, 164
Practices, social, 270, 274-275
Pragmatic implication. See
Contextual implication
Predicative sentences, 157
Predictives, 36, 42
Preparatory conditions, 56, 128, 131
Preparatory rules, 129
Presupposition, 102, 137, 155,
297n12
pragmatic, 158-165
semantic, 155-158, 298n21
Principle of charity (PC), 168
Prohibitives, 47, 49
Promises, 36, 50-51, 56, 75, 128,
129, 130
Pronouns, 24, 25, 27-28,32
Proper names, 25, 28, 31-32
Propositional content, 11, 24, 34, 36,
39,47,49-50,57,71,123
conditions, 56, 128, 131
rule, 129
stage, 241
Propositions, 9, 44
Pseudo-imperatives, 177
Puns, 5, 97, 98
Quality, presumption of (QL), 63,
64, 106, 107, 167
Quantity, presumption of (QT) , 65,
107, 167
Queclaratives, 177
Questions, 36, 47, 48, 63, 186
Quine, W., 237
Rawls, J., 293n1
Reciprocity of perspectives, 268
Reciting, 97, 99
Redundancy rules, 141
Reference, 9, 19, 20, 25-28, 31
generic, 26
nonspecific, 26
 Index
Reference (continued)
potential, 31
specific, 26
Referential conditions, 155
Referential intent, 25
Referential opacity, 29, 286n8
Referential transparency, 29
Referring expressions, 24, 38
Reflexive paradox, 15
Reflexive (R-)intentions, 15, 39, 40,
46, 54, 57, 58, 59, 84, 85, 95, 96,
97,98,101,102, 110, 117, 118,
123, 124, 125, 126, 127, 132, 151,
173,197,224, 290n6, 294nl,
294n2
Relation, 24, 167
Relevance, presumption of (RE) , 63
Representatives, 41
Requestions, 177
Requestives, 47, 48, 198
Requests, 10, 16, 36, 63, 73, 75, 76,
80, 85-86, 199
Requirements, 47, 48-49
Responsives, 43, 46, 103
Retractives, 43, 45
Rewrite rules, 144
Rhetic act, 19-20
R-intentions. See Reflexive intentions
Role negotiation, 104
Roles, 278-279
Role standards, 229
Rosenberg, J., 149
Ross, J., 209, 300n29
Rule categories, 277-279
Rule-condition connection (R-C),
128, 129, 130
Rules
constitutive, 127-131
and conventions, 120-122
social, 94-95, 275-277
collective, i75, 276
cooperative, 275-276
coordinative, 275
regulations, 275, 276-277
Russell, B., 32
Sadock, J., 70,80, 174, 175-177,
180-183, 185, 186, 198, 209-210,
221, 222, 230, 301nl, 301n2,
301n4, 301n8, 304nl
Saliency, 80-81
Sarcasm, 67-68, 71, 130, 292n8
View publication stats
326
Saying, 8-10
Saying that, 29-34
Scheff, T., 268-269
Schelling, T., 81, 285n4
Schiffer, S., 133, 137, 152, 153, 154,
205-206, 286n12, 286n13, 289nl,
293n5, 296nll, 309nl
Schreiber, P., 221, 305n17
Schweller, K., 261, 265-266
Scope shifting, 231, 232
Search procedure, 91
Search strategies, 93
Searle, J., 11, 13-14,40,41,47,55,
56, 70, 80, 127-131, 150-151,
153, 154, 177, 184, 186, 187-188,
189, 190-191, 192, 197, 288nll,
289nl, 289n4, 291n3, 293n5,
293nl, 293n2, 295n3, 295n4,
295n5, 302n12, 302n13
Secord, P., 271, 277, 278
Semantic properties, 140
Semantic relationships, 98, 138-139,
144-145, 146, 149, 156-157
Semantic representation, 229, 239,
244-245
Semantic rules, 127, 130
Sense, 19, 24, 31
Sentence type, 34
and type of saying, 24-25
Sequencing, presumption of (SE),
63
Should, 198
Silver, M., 253
Sincerity, presumption of (SI), 39,
44, 57, 63, 161
Sincerity condition, 128, 131
Sincerity rule, 129
Singer, M., 261
Slips of the tongue, 33, 134, 231
Small talk, 99
"Sneaky presupposition," 101,
102
Social collectivities, 277-283
Social concepts, 267-284 passim
Social groups, 277, 280, 281
"Socially real," 270
Social moves, 103-107
Social regularities, 121, 270-277
Social standard, 121, 273
Social types, 278, 279
Society, 267-268
Speaking idiomaticallv 188-189
Index 327
Speech act schema (SAS), 4-12,. Transformations, 140, 144
20-38, 60-107 passim, 234-266 Truth
passim conceptual,68
and the analysis of linguistic obvious factual, 68
communication, 85-89 Truth conditions, 9
and the determination of reference, Tyler, L., 254
25-28
elaborated version of, 76-81 Understatement, 66, 67
explanatory value of, 89-91 Uptake, 4,13, 17,70,103,151,294
and literal indirect acts, 71, 73-76 Urmson, J., 300
and nonliteral indirect acts, 71-76 Utterance act, 3, 17, 285nl, 285n5
perlocutionary effects generated
from, 17-18 Vacuous terms, 32
psychological reality of, 91-93 Valian, V., 307
and psychology, 234-266 passim Vendler, Z., 289nl, 304n4
simple version of, 4-8 Verbal taboo, 100
and social psychology, 93-96 Verdictives, 40, 108, 110-113
Speech comprehension, 240-242
illocutionary force stage, 260-266 Warning, 49, 71
operative meaning stage, 242-257 Weighted principle of charity (WPC) ,
propositional content stage, 168
257-260 Weiser, A., 101, 102, 107
Speech production, 236-240 Whimperatives, 176, 177-180,
Speech production model (SP), 238 182-183, 184-187, 196, 201
Spira, J., 254-255 Wh-interrogatives, 25
Spoonerism, 86 Will, 198
Standardization thesis, 192-195 Williams, F., 238
Standardized indirection, 195-196 Wilson, D., 155, 158, 299n24
Statistical regularity, 121, 272 Word order, 36
Stereotype, 279 Would, 197-198
Storytelling, 97, 98
Strawson, P. F., 4, 13, 117, 131, Yes/No interrogatives, 25
132, 145-146, 152, 155, 156,
293n5, 298n14, 298n16, 300n29 Ziff, P., 296n9
Success conditions, 55
Suggestives, 36, 43-44, 45, 46, 182
Superlatives, 67
Suppes, P., 292n2
Suppositives, 44, 45, 46
Surrendering, 36, 51
Synecdoche, 68, 292n7
Synonymy, 140, 143, 149
Syntactic liberties, 199, 228-233
Tag questions, 177
Tannenbaum, P., 238
Temporal descriptions, 24
Tense, 24
Thank, 52,,54
Thomson, J., 285n2
Time, 24, 36
Transformational grammar, 191-192